fix(audit): handle Redis read timeouts

[man4ish]

Fix Redis read timeout crash loop

Problem

The security-audit worker could terminate when the blocking Redis
XREADGROUP operation reached the redis-py socket read timeout while the
stream was idle.

Docker would restart the worker, causing a recurring crash loop approximately
every five seconds.

Fix

Handle the expected Redis read timeout at the worker's outer blocking-read
boundary so an idle stream does not terminate the worker process.

Unexpected exceptions remain distinguishable and are not silently swallowed.

Regression coverage

Tests verify:

• read timeout does not terminate the worker
• the read loop continues
• messages are still processed after a timeout
• existing ACK behavior remains intact
• unexpected exceptions retain their existing behavior
• clean shutdown remains functional

Validation

• focused worker tests: 12 passed
• full test suite: 179 passed
• runtime idle-worker validation: PASS (stable across 6 x 5s idle read cycles, 0 restarts)
• subsequent event processing: PASS (new event consumed and persisted, 0 duplicates)
• no crash loop observed
• git diff --check: PASS

Scope

This PR only fixes the worker's Redis read-timeout handling.

No Interaction implementation is included.
No API Gateway changes are included.
No Control Center changes are included.
No authentication/session changes are included.