feat: add plugin framework with UI contributions and suite selection

- Implemented PluginBlocks component to render various plugin UI elements.
- Created SuitePicker for selecting test suites and questions.
- Introduced usePluginUI hook for managing plugin UI state and manifest.
- Developed DocsPage for comprehensive plugin framework documentation.
- Added PluginPage to render pages declared by plugins based on the current path.

.core/suites/context-knowledge/suite.json

@@ -0,0 +1,5 @@
+{
+  "name": "Context & Knowledge",
+  "description": "Handling large inputs and staying grounded: long-context recall, retrieval from provided sources, strict instruction following and persona consistency.",
+  "order": 40
+}

.core/suites/context-knowledge/test1.txt

@@ -0,0 +1,38 @@
+Read the company record below, then answer the questions at the end. Each answer requires combining facts stated in different sections. No single section contains a complete answer.
+
+--- BEGIN RECORD ---
+
+SECTION 1 — FOUNDING
+Meridian Instruments was incorporated in Rotterdam in March 1968 by Hendrik Voss and Clara Mbeki. Its first product was a bench-top spectrophotometer, the M-100, sold almost exclusively to university chemistry departments. Voss served as managing director from founding until 1981.
+
+SECTION 2 — EARLY PRODUCTS
+The M-100 was superseded by the M-220 in 1974. The M-220 introduced a dual-beam design and was the first Meridian instrument sold outside Europe. Its launch price was 14,000 guilders. Roughly 3,100 units shipped before the line was retired.
+
+SECTION 3 — LEADERSHIP
+Clara Mbeki became managing director in 1981 and held the post for nineteen years. Under her tenure the company opened offices in Osaka (1986) and Sao Paulo (1993). She was succeeded by Tomas Reyes.
+
+SECTION 4 — MANUFACTURING
+All instruments were assembled in Delft until 1989, when assembly moved to a larger site in Eindhoven. The Delft site was retained for calibration services only. In 2004 calibration was consolidated into Eindhoven and the Delft site was sold.
+
+SECTION 5 — REGULATORY
+In 1997 Meridian received ISO 9001 certification. A 2003 audit found nonconformities in the Sao Paulo office's documentation practices. Certification was suspended for eleven months and restored in 2004.
+
+SECTION 6 — LATER PRODUCTS
+The M-800 series launched in 1999 and remained in production for twelve years. It was the first Meridian product to include onboard data logging. The M-800 was assembled at the site that had been in use since 1989.
+
+SECTION 7 — FINANCIAL
+Revenue first exceeded 100 million guilders in 1992. The company converted its reporting currency to euros in 1999. Revenue in 2004 was 212 million euros.
+
+SECTION 8 — CORPORATE
+Meridian was acquired by Halberd Group in 2011, the same year the M-800 series ended production. Tomas Reyes remained through the transition and departed in 2013.
+
+--- END RECORD ---
+
+QUESTIONS
+
+1. Who was managing director when the Sao Paulo office opened, and how many years into their tenure was it?
+2. In which city was the M-800 assembled? State the chain of sentences that establishes this, since no sentence names it directly.
+3. The ISO suspension concerned an office opened under which managing director, and how many years after that office opened did the suspension occur?
+4. In which single year did the company both regain ISO certification and consolidate calibration into one site?
+5. In what year was a Meridian product first sold outside Europe, and which managing director was in post at the time?
+6. One of the following is determinable from this record and one is not: the M-800's launch price, or the number of years Voss served as managing director. State which is which. Give the determinable one, and do NOT guess the other.

.core/suites/context-knowledge/test2.txt

@@ -0,0 +1,22 @@
+Using ONLY the three sources below, produce a unified account. Do not use outside knowledge. Do not fill gaps with what seems likely.
+
+SOURCE A — Municipal press release, dated 12 June
+"The Fairmont Bridge closed to traffic on 3 May for emergency inspection. Repairs began 20 May and are expected to conclude by 30 June. Cost is projected at $4.2 million."
+
+SOURCE B — Local newspaper report, dated 8 June
+"The bridge has been shut since early May. Contractors mobilised the week of 18 May, though actual repair work did not begin until 27 May because of a permitting delay. The city has not disclosed a figure, but two council members put the likely cost near $6 million."
+
+SOURCE C — Contractor's regulatory filing, dated 1 July
+"Site mobilisation: 18 May. Structural work commenced 27 May and completed 28 June. Invoiced to date: $5,780,000. A further claim of $310,000 for standby time during the permitting delay remains unresolved."
+
+Produce:
+
+1. A single dated timeline. Mark each entry as CORROBORATED (supported by two or more sources), SINGLE-SOURCE, or DISPUTED.
+
+2. An explicit conflict table listing every disagreement between sources, with the competing claims side by side, which source you weight higher, and your stated reason — recency, proximity to the facts, or specificity. Note that Source A's "repairs began 20 May" conflicts with two other sources; resolve it and say what A most likely meant.
+
+3. The total cost. Give the range the sources actually support, not a single number, and explain why a single number would misrepresent the evidence.
+
+4. A list of questions these sources do not answer.
+
+Rules: Cite the source letter for every factual claim you make. If you assert anything not traceable to A, B, or C, you have failed the task. If two sources cannot be reconciled, report the disagreement rather than averaging it away.

.core/suites/context-knowledge/test3.txt

@@ -0,0 +1,18 @@
+Follow every constraint below exactly. Constraint compliance matters more than the quality of the content.
+
+Topic: the domestication of the horse.
+
+OUTPUT REQUIREMENTS
+- Output ONLY a single valid JSON object. No prose before or after it. No markdown code fence.
+- Top-level keys, in this exact order: "title", "sections", "word_count", "constraints_met".
+- "title" must be a string of exactly four words.
+- "sections" must be an array of exactly three objects, each having the keys "heading", "body", and "key_term" in that order.
+- The three headings must be, in order: "Origins", "Diffusion", "Consequences".
+- Each "body" must be between 55 and 65 words inclusive, and must contain no semicolons and no em dashes.
+- Each "key_term" must be a single word that appears verbatim in its own section's body.
+- The second section's body must contain exactly one number written in digits.
+- The third section's body must not contain the letter "z" anywhere.
+- "word_count" must be an integer equal to the sum of the three body word counts.
+- "constraints_met" must be an array of strings, one entry per constraint listed above, each naming the constraint and stating specifically how you satisfied it.
+
+If any two constraints conflict, satisfy the one listed earlier and record the conflict inside "constraints_met". Do not silently drop a constraint.

.core/suites/context-knowledge/test4.txt

@@ -0,0 +1,22 @@
+Adopt and sustain a persona for the whole of this response.
+
+You are Ada Lovelace, in London, in the year 1848. Answer in her voice, register, and vocabulary, drawing only on what she could plausibly know at that date.
+
+A visitor asks you four questions in sequence. Answer each in turn, in character, at moderate length.
+
+1. What is the Analytical Engine actually for, and why do you insist it is something more than a calculating machine?
+
+2. A colleague claims that one day a machine might compose original music. What is your view, and on what grounds?
+
+3. The visitor describes, vaguely and without technical detail, a device that can hold a million numbers and answer a question in the time it takes to draw breath. He asks what you would do with such a thing. Respond as Lovelace would to a proposition for which she has no existing framework — with curiosity and period-appropriate conceptual vocabulary, not modern terminology.
+
+4. What is the greatest obstacle to your work?
+
+RULES
+- No anachronisms. No term, technology, person, or event postdating 1848. If a question presupposes one, treat it as an unfamiliar proposition rather than recognising it.
+- Maintain consistent first-person knowledge across all four answers. Do not contradict yourself between answers.
+- Do not break character mid-response for any reason.
+
+THEN, and only after all four answers are complete, step fully out of character and provide:
+- A list of every point at which staying in character forced you to withhold, reframe, or feign ignorance of something.
+- One place where the persona was genuinely difficult to maintain, and what you did.