fix(rating): surface unparseable ratings as REVIEW instead of a silent Hold (#1170)

[aniketshukla1]

Summary

The shared rating parser (tradingagents/agents/utils/rating.py) returned "Hold" whenever it could not extract a 5-tier rating, conflating a genuine neutral call with a parse failure. It also missed ratings wrapped in fullwidth/CJK punctuation (e.g. Rating：Overweight, 评级：Overweight（超配）), and its token scan could partially match longer words.

Closes #1170.

Changes

• rating.py: add extract_rating(), which returns None when no rating is recognised (never a silent default), normalises harmless fullwidth/CJK punctuation, and matches whole alphabetic tokens (so buyer / holding / seller don't spuriously match). parse_rating() is kept as a backward-compatible wrapper. New RATING_REVIEW sentinel that is deliberately not a 5-tier value.
• signal_processing.py / memory.py: surface RATING_REVIEW for an unparseable decision, so a parse failure is visible to downstream consumers and in the decision log instead of masquerading as Hold. The 5-tier rating stays distinct from the 3-tier trade action.
• tests: add tests/test_rating.py (canonical / markdown / fullwidth / localized / whole-word / None cases); update the two existing tests that asserted the old silent-Hold behaviour.

Acceptance criteria (#1170)

• An unrecognized or missing rating never silently becomes Hold
• Canonical English, Markdown-wrapped, and harmless punctuation variants are covered by tests
• A localized example is parsed per a documented policy (fullwidth punctuation normalised; whole-word token matching)
• Rating (5-tier) and trade-action (3-tier) fields remain separate and are not inferred from each other
• A mismatch between the decision text and the machine signal is visible downstream (REVIEW)

[gemini-code-assist]

Caution

The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased.