refactor: Unify return types and casing

[adnanrhussain]

Please review this PR by each commit. This PR primarily refactors and unifies a consistent result type across the evals and ensure the metadata is accurate

1. 3a0b96c235505ca92ac324269bf05c741b7f82d0

refactor(types): unify TextComplexityLevel and tighten evaluator return types

• Rename ComplexityLevel → TextComplexityLevel; all evaluators now share a single snake_case enum (slightly/moderately/very/exceedingly_complex)
• Remove VocabularyComplexityLevel (lowercase spaced) in favour of unified enum; complexity_score now uses snake_case values
• Add COMPLEXITY_LEVEL_LABELS map for display strings (e.g. 'Slightly complex')
• Rename internal types: VocabularyComplexity → VocabularyInternal, GradeLevelAppropriateness → GradeLevelAppropriatenessInternal
• Move SentenceStructureInternal to its schema file and export it
• Tighten EvaluationResult TScore: string → TextComplexityLevel (vocab, SS) and string → GradeBand (GLA)
• Redesign TextComplexityEvaluator to return TextComplexityResult map instead of a nested EvaluationResult wrapper; each key holds the sub-evaluator result or { error: Error } directly
• Remove buildCombinedReasoning; callers access per-evaluator reasoning directly from the result map
• Make runSubEvaluator generic to preserve TScore/TInternal types through the p-limit boundary
• normalizeLabel normalises LLM output variations to canonical snake_case values; returns TextComplexityLevel | null (no cast needed)
• Update unit and integration tests for new enum values and result shape

2. 14b91d1c0c08c52e781acd5fab7545156bccf752

3. Removed unused exports and metadata, like promptVersion and timestamp

4. Fixed the telemetry endpoint

[codecov]

Codecov Report

❌ Patch coverage is 80.95238% with 4 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...ript/src/evaluators/grade-level-appropriateness.ts	50.00%	1 Missing ⚠️
...ks/typescript/src/evaluators/sentence-structure.ts	83.33%	1 Missing ⚠️
sdks/typescript/src/evaluators/text-complexity.ts	88.88%	1 Missing ⚠️
sdks/typescript/src/evaluators/vocabulary.ts	66.66%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

[czi-fsisenda]

Looks good!
Just one comment that is unrelated to the primary change in this PR so you can address it here or later.

[czi-fsisenda]

I think updates to the prompt should result in us creating a new prompt file. The convention could be [eval_id]_[eval_ver]/user.txt or [eval_id]/user_[eval_ver].txt
Updating the prompt version without updating the eval version misrepresents the evaluator. Updating both seems redundant. Having a prompt version seems to just introduce the opportunity to mismatch evaluator and prompt versions.