Cleanup serving stale while origin server down

[masaori335]

HttpTransact::HandleCacheOpenReadHit try to return stale contents by checking is_server_negative_cached(s), but actually, this is dead code. Because when all origin servers are marked down, HttpTransact::OSDNSLookup serves stale contents.

This PR marked as draft because new AuTest depends on #13069. (update: this is ready)

[Copilot]

Pull request overview

This PR removes unreachable stale-serving logic in HttpTransact::HandleCacheOpenReadHit() and relies on the existing origin-selection/DNS path for serving stale when origins are unavailable, with a new AuTest to cover the “origin down” case and a small documentation update.

Changes:

• Remove dead code in HttpTransact::HandleCacheOpenReadHit() related to origin negative-caching stale serve.
• Add a new cache replay test validating stale serving while the origin is marked down, and register it in the cache gold test.
• Expand the proxy.config.http.down_server.cache_time documentation to mention stale serving behavior.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

File	Description
src/proxy/http/HttpTransact.cc	Removes unreachable origin-negative-cache stale-serve branch; keeps parent-down stale behavior.
tests/gold_tests/cache/replay/proxy_serve_stale_origin_down.replay.yaml	New replay that caches an object, marks origin down, verifies stale serve, then revalidation.
tests/gold_tests/cache/proxy_serve_stale.test.py	Registers the new replay under the existing cache stale-serving gold test.
doc/admin-guide/files/records.yaml.en.rst	Adds explanatory text for down-server cache time behavior.