fix(frontend): stabilize label-set workflows

[angelo-lacson]

Summary

Fixes the misleading state and validation failures encountered while creating labels and assigning a label set during corpus creation.

Fixes #2295.

Changes

• Normalize label colors to a backend-compatible #RRGGBB value while keeping native color inputs stable for stored values with or without #.
• Model and inspect the nested createAnnotationLabelForLabelset mutation payload, report backend failures accurately, and await refreshed label-set data before closing the create form.
• Keep Add Label beside search for populated text, document, relationship, and span label lists.
• Return the selected LabelSetType from LabelSetSelector so controlled consumers such as CorpusModal immediately display the user's selection.
• Add component-test coverage for rejected create mutations and controlled label-set selection.
• Add a changelog.d fragment.

Test plan

• NODE_OPTIONS=--max-old-space-size=4096 yarn build via the frontend Docker builder — passed (TypeScript and Vite production build).
• Prettier check for all changed frontend and test files — passed.
• python3 scripts/collate_changelog.py --check — passed.
• Added focused Playwright component tests; the browser component suite was not executed locally.

Checklist

• Tests pass locally for any code this PR touches (focused coverage added; browser component suite not run locally)
• pre-commit run --all-files passes (not run locally)
• TypeScript compiles cleanly (yarn build or yarn lint), if frontend code changed
• A changelog fragment was added under changelog.d/ for the user-facing fixes
• No new dependency was introduced

Contributor License Agreement

By submitting this pull request, I agree to license this contribution under the project's Contributor License Agreement.

[codecov]

Codecov Report

❌ Patch coverage is 18.18182% with 18 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...nd/src/components/labelsets/LabelSetDetailPage.tsx	0.00%	12 Missing ⚠️
...d/src/components/widgets/CRUD/LabelSetSelector.tsx	33.33%	4 Missing ⚠️
frontend/src/components/corpuses/CorpusModal.tsx	33.33%	2 Missing ⚠️

📢 Thoughts on this report? Let us know!

[JSv4]

Thanks for this — nice cleanup. Can you just make sure the non-Claude actions pass and I'd be happy to merge this?

1. Lint / Check any-baseline — needs a one-line follow-up from you

Your changes removed two explicit anys, which is exactly the right direction, but the ratchet gate fails when the count drops without frontend/.any-baseline.json being lowered to match:

[FAIL] `any` count dropped by 2 but the baseline file was not updated.
       Run `yarn any:write` and commit the result so the gate ratchets down.

From frontend/:

yarn any:write
git add .any-baseline.json && git commit -m "chore(frontend): ratchet any-baseline"

Background on the gate is in docs/frontend/any-baseline.md (issue #1448). This is also why Build, Unit Tests, Component Tests and Docker all show skipping — they're needs: lint, so they never start until Lint is green. One commit should turn the whole column over.

2. claude-review — not your fault, please ignore

Fork-based PRs don't receive repository secrets or an OIDC token, so the review action has nothing to authenticate with. The error it prints (Did you remember to add id-token: write to your workflow permissions?) is misleading — the permission is declared in the workflow; GitHub drops it for fork-triggered runs by design, and no change on your side can restore it.

We're patching the workflow to skip cleanly on forks rather than report a red X you can't act on. Nothing for you to do here.

[JSv4]

Sorry wanted to merge this. See PR #2328