Fix #4007: lab-ops-hub: Worker-fallback extractor — complete HTML character-referen #5115

Implement GitHub issue #4007 in developing-today/code: lab-ops-hub: Worker-fallback extractor — complete HTML character-reference decoding parity

Problem

The lab-ops-hub Worker-fallback extractor (pkgs/lab-ops-hub/src/lib.ts, decodeHtmlEntities) only decodes a small named-entity table and requires a trailing semicolon, so its output diverges from the container Extractor (Python html.parser + html.unescape) on standard references — stored titles/text then depend on which crawling path ran.

Found by Codex review round 4 on PR #3918 (comment 4437512013), dispositioned there as a follow-up because the fix (complete HTML5 named-reference data + semicolon-omission rules) is larger than a review-fix round. Sibling of #569.

Evidence (verified against the container Extractor, Python 3.13.13)

Input: <title>&Alpha; &AMP;</title><p>&rarr; &#65 &amp B</p>

• Container: title Α &, text → A & B
• Worker fallback: every reference preserved literally

Scope

• Decode the complete HTML5 named character references (the data behind Python's html.entities.html5), including the legacy semicolon-less forms and the spec's semicolon-omission rules for text content (e.g. &#65 -> A, &amp -> & in the cases html.unescape expands).
• Keep the existing numeric-reference normalization (C1 remap, U+FFFD rules) intact.
• Add shared parity fixtures (test/extractor-fixtures.json) pinning the cases above plus a spread of named references; expected values verified against the container suite before pinning. Both suites (node + python) must stay green.

Acceptance

• The example above produces identical {title, links, text} from the Worker fallback and the container Extractor.
• New fixtures cover: a non-Latin-1 named reference (→), an uppercase legacy reference (&), a semicolon-less numeric reference, and a semicolon-less named reference.

Acceptance criteria

• The Worker-fallback extractor (pkgs/lab-ops-hub/src/lib.ts, decodeHtmlEntities) decodes the complete HTML5 named character references, including legacy semicolon-less forms, with the spec's semicolon-omission rules for text content matching Python html.unescape.
• Filed evidence case: input <title>&Alpha; &AMP;</title><p>&rarr; &#65 &amp B</p> yields title Α & and text → A & B from the Worker fallback, identical to the container Extractor.
• A differential test corpus (named references with/without semicolons, numeric references with/without semicolons, ambiguous ampersands in text and in attribute-like contexts) matches html.unescape output case for case.
• Existing extractor test suites pass; no change to the container Extractor.

Rules from the repo owner: (1) Deliver the finished work as a pull request in this run; put "Fixes #4007" in the PR description so the issue closes on merge. (2) Cover the whole set in the one PR: if the issue names a set (files, subcommands, cases), fix every member, and batch any directly related dependency updates into the same PR — never one PR per item or per dependency. (3) Run the relevant tests and checks before opening the PR. (4) Merging is permitted but never required: merge only when checks pass and you are reasonably certain the PR is correct and low-risk — otherwise leave it open, and state your confidence and what a human should check. (5) If you hit a genuine blocker (missing credentials, a decision only the owner can make), stop and say exactly what is needed in your final summary instead of guessing.

Launched by pm-dispatch for developing-today/code. Queue task and OpenCode session ids follow in a comment below.