fix(preprocess): recover markdown-wrapped ticker hints

[keli-wen]

Summary

• Normalize Markdown link and */** emphasis decoration on a scan-only copy before extracting exchange-qualified ticker hints.
• Preserve stored Markdown, content_hash, and document identity semantics.
• Add deterministic markup fixtures, including the dominant multiline (NASDAQ:\n\n[PODD](...)) representation, plus a bounded live recall control that samples up to 25 PR Newswire articles with an independent oracle.
• Keep multi-symbol list grammar and exchange-whitelist expansion out of scope; follow-up discussion is tracked in bug: exchange-qualified multi-symbol lists are only partially captured #108.

Root Cause and Controls

The extractor tests only covered ideal plain text while the article boundary supplies cleaned Markdown. The existing live verifier stopped after RSS and listing discovery, so it never exercised article representation or ticker-hint recall.

This change adds two executable controls: an offline fixture gate requiring full recall for single-line links, multiline links, bold, and italic forms without mutating stored content; and a daily live gate requiring 100% recall for the first supported exchange-coded symbol in each sampled group. A successfully parsed sample with zero supported symbols reports SKIP and passes neutrally, avoiding false failures on thin-volume windows.

Related Issue

Closes #107

Verification

• bash scripts/verify.sh — 291 tests passed; 88.55% coverage; format, lint, typing, and import contracts passed.
• python scripts/verify_news_e2e.py — 1,102 observations discovered; 25/25 articles parsed; 15/15 expected ticker hints recovered (100% recall).
• python examples/preprocess/01_news_pr_wire.py — stored Markdown retained and CCL extracted.

Checklist

• The title uses English Conventional Commit format: type(scope): summary.
• The related issue or design discussion is linked when applicable.
• bash scripts/verify.sh passes.
• Every applicable live component gate passes, or this PR states why none applies.
• Public behavior has focused tests, an example, and documentation where applicable.
• The PR is complete, small, and contains no unrelated changes.