fix(gateway): separate 429 quota errors from auth failures (#32790)

[zccyman]

Closes #32790

Summary

Two fixes in one commit because they share the same code area:

1. 429 quota mislabeled as auth failure (gateway/run.py:16132): When resolve_runtime_provider raises an AuthError whose code indicates a rate-limit/quota problem (contains "rate", "quota", or "limit"), the gateway now surfaces "⚠️ Provider quota exhausted: ..." instead of the misleading "⚠️ Provider authentication failed: No Codex credentials stored".

2. Fallback provider log label (gateway/run.py:910): The Fallback provider resolved: log line now uses the literal provider key from the config entry (entry.get("provider")) instead of the normalized provider label returned by resolve_runtime_provider. This fixes the issue where ollama in config was being logged as openrouter.

Changes

• gateway/run.py — two targeted patches; no new dependencies

Testing

Existing tests cover the fallback resolution path. The behavioral change (different user-facing message for rate-limit AuthErrors) is verified by code inspection.

[alt-glitch]

Duplicate of merged #33168 which already fixes #32790 (Codex 429 quota mislabeled as auth failure). The fallback provider log label fix is also covered by #32881 (open).