fix: preserve work intent when normalizing permissions

[giscebot]

Summary

• Treat work_intent as canonical when normalizing intent-classifier output.
• Normalize inconsistent write_permission values to match work_intent instead of downgrading work_allowed to review_only.
• Expose normalization_warnings in intent-classifier metadata when a field is corrected.

Why

A trusted PR comment asking @giscebot to add an e2e test produced raw LLM output with work_intent=work_allowed and write_permission=none. The previous normalizer silently converted that to review_only, making the dashboard look like the LLM had classified the request as read-only.

Tests

• pytest tests/test_intent_classifier.py tests/test_parser.py

[pilipilisbot]

Approved. I reviewed the intent-classifier normalizer, the enqueue path that applies the LLM result, and the dispatch path where work_intent controls write/push allowance, routing, and timeouts. Treating work_intent as canonical here matches the existing operational gate, and normalizing write_permission upward for work_allowed avoids silently misreporting an implementation request as review_only while preserving an audit trail through normalization_warnings.

Checked locally in an isolated temp virtualenv: /tmp/gab-pr149-venv/bin/python -m pytest tests/test_intent_classifier.py tests/test_parser.py (28 passed).