What this PR does / why we need it:

The metrics HTTP server runs as a thread in the Gunicorn master, which forks the workers. Its default WSGIRequestHandler writes an access-log line to stderr for every scrape. A fork while that thread holds the stderr buffer lock leaves the new worker with the lock held forever: the worker blocks on its first log line ("Booting worker") and never serves a request. Details, stacks and a reproduction in #6928 (same class of problem as #6647).

This adds _QuietWSGIRequestHandler (no-op log_message, like prometheus_client's own _SilentHandler) and uses it for both the IPv4 and the dual-stack server built by _make_metrics_httpd. Scrapes are no longer written to stderr; nothing else changes.

Tested on Kubernetes with --max-requests 20 and a scrape storm: 5 hangs in 318 forks before, 0 hangs in 2,000 forks after. Locally (steps in #6928): a hang after 12–29 forks before, none in 336 forks after.

Which issue(s) this PR fixes:

Fixes #6928

Checks

• I've made sure the tests are passing. (pytest sdk/python/tests/unit/test_metrics.py: 117 passed; ruff check and ruff format --check pass; mypy feast reports no errors in metrics.py)
• My commits are signed off (git commit -s)
• My PR title follows conventional commits format

Testing Strategy

• Unit tests – TestMetricsHttpdDoesNotLog: a request to the server built by _make_metrics_httpd writes nothing to stderr (fails before the change, passes after)
• Integration tests

🤖 Generated with Claude Code