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/language/suite.json

@@ -0,0 +1,5 @@
+{
+  "name": "Language",
+  "description": "Foundational understanding and generation: sentiment, entities, similarity, coreference, summarization and translation.",
+  "order": 10
+}

.core/suites/language/test1.txt

@@ -0,0 +1,10 @@
+Analyze the sentiment of the following customer review.
+
+Review:
+"I ordered the noise-cancelling headphones after reading three glowing write-ups. The build quality genuinely surprised me — the hinges feel solid and the earcups are the most comfortable I've worn. Battery life is as advertised. That said, the companion app crashed twice during setup, and support took nine days to answer a one-line question. I still reach for these every morning."
+
+Do all of the following:
+1. State the dominant overall sentiment (positive, negative, neutral, or mixed) in one word.
+2. Give an aspect-level breakdown: for each of build quality, battery, software, and customer support, label the sentiment and quote the span of text that justifies it.
+3. Identify the single sentence that most complicates a naive positive/negative classification, and explain why.
+4. Rate your confidence in the overall label from 0 to 100 and justify the number.

.core/suites/language/test2.txt

@@ -0,0 +1,11 @@
+Extract every named entity from the news excerpt below and classify it.
+
+Excerpt:
+"Reuters, Nairobi — On 14 March 2023, the Kenya Wildlife Service and the International Union for Conservation of Nature signed a five-year agreement at the Serena Hotel to expand rhino corridors across Tsavo East National Park. Dr. Amina Okello, who has led KWS since 2019, said the deal was modeled on a program Botswana abandoned in 2017. The World Bank will contribute $42 million; the remainder comes from the Leakey Foundation. Reporting by Jonah Mwangi; editing by Priya Raghunathan."
+
+Return a markdown table with the columns: Entity | Type | Justification. Use only these types: PERSON, ORG, LOCATION, DATE, MONEY, FACILITY.
+
+Then answer separately:
+1. Which mention is an alias for an entity already listed under a different surface form?
+2. Which single string is genuinely ambiguous between two types, and why?
+3. Which entity is a nested entity — one entity name contained inside another — and how did you handle it?

.core/suites/language/test3.txt

@@ -0,0 +1,18 @@
+Rank the following four phrases by how close each is in meaning to the reference phrase, from closest to most distant.
+
+Reference: "The committee postponed the vote."
+
+Candidates:
+A. "The vote was delayed by the committee."
+B. "The committee cancelled the vote."
+C. "The council deferred the ballot."
+D. "The committee voted to postpone other business."
+
+For each candidate:
+- Give a similarity score from 0.00 to 1.00.
+- Give one sentence justifying the score.
+
+Then answer:
+1. Why are B and D traps? For each, name the specific surface feature (shared vocabulary, shared syntax) that makes it look closer than it actually is.
+2. C shares almost no words with the reference. Explain why lexical overlap is a poor proxy for meaning here.
+3. A is a voice alternation of the reference. Is it fully equivalent, or does the passive shift emphasis in a way that matters? Defend your answer.

.core/suites/language/test4.txt

@@ -0,0 +1,13 @@
+Resolve every referring expression in the passage below.
+
+Passage:
+"The trophy would not fit in the brown suitcase because it was too large. Marta handed it to Dylan anyway, and he wedged it under the seat. When the conductor asked about it, she said it belonged to her sister, who had won it in Lisbon the year before. He did not believe her."
+
+For each pronoun and definite noun phrase, state:
+- The referent.
+- The single strongest cue that determines it (syntactic position, world knowledge, discourse recency, or gender agreement).
+
+Additional requirements:
+1. The first "it" is a classic Winograd case. State the referent, and then state what the referent would become if "large" were changed to "small" — and explain what that swap demonstrates about how the ambiguity is resolved.
+2. Where a reference is genuinely ambiguous, say so and give both readings. Do not silently pick one.
+3. Identify which single referring expression in this passage is the hardest to resolve, and why.

.core/suites/language/test5.txt

@@ -0,0 +1,14 @@
+Summarize the passage below twice, under strict and different constraints.
+
+Passage:
+"Mycorrhizal networks are symbiotic associations between fungi and plant roots, occurring in roughly ninety percent of land plant families. The fungus extends hyphae far beyond the root zone, increasing the plant's absorptive surface area by one to two orders of magnitude, and delivers phosphorus, nitrogen, and water. In exchange it receives carbon fixed during photosynthesis, between four and twenty percent of the plant's total photosynthate. Because a single fungal network can connect multiple host plants, researchers have documented transfers of carbon and defensive signaling compounds between neighboring trees. Popular accounts compress this into the phrase 'wood wide web,' which several soil ecologists argue overstates the evidence: most transfer studies were conducted in greenhouses, effect sizes under field conditions are small and inconsistent, and the claim that mature trees preferentially nourish their own seedlings rests on a handful of studies that have not been independently replicated."
+
+Produce:
+1. An EXTRACTIVE summary: exactly three sentences, copied verbatim from the passage, chosen so that both the mechanism and the scientific caveat survive.
+2. An ABSTRACTIVE summary of exactly 40 words. Not 39, not 41. Reworded throughout, readable by a non-specialist.
+
+After each, state its word count.
+
+Then answer:
+- Name one fact that survived both compressions and one that survived only one, and explain the tradeoff you made.
+- The passage contains a claim and a challenge to that claim. Which summary handles the tension better, and why?

.core/suites/language/test6.txt

@@ -0,0 +1,17 @@
+Translate the passage below into French and into Japanese.
+
+Passage:
+"Look, I'm not going to beat around the bush. The launch was a train wreck. Marketing dropped the ball, engineering was flying blind, and by the time anyone pulled the plug we were deep in the red. Still, we live and learn."
+
+Requirements:
+- Preserve the register. This is a blunt, informal remark from a manager to peers in a closed meeting. It is not a press release, and it is not crude.
+- Do not translate the idioms literally. For each idiomatic expression, use the natural equivalent in the target language.
+- Japanese must use an appropriate level of formality for this setting, and you must state which level you chose and why.
+
+After each translation, provide a table with the columns:
+English idiom | Your rendering | Literal back-translation | What was lost or gained
+
+Then answer:
+1. Which idiom has the weakest equivalent in each target language, and how did you compensate?
+2. "In the red" is an accounting metaphor. Does the color metaphor survive in both languages? Say what each language does instead if it does not.
+3. Which is harder to carry across, the idioms or the register? Defend your answer.