fix(relay): tolerate bounded Redis fsync stalls

[BeelixGit]

Summary

• raise only the pooled Redis admission operation response bound from redis-rs's 500 ms default to 3 seconds
• preserve fail-closed rate limiting and the atomic Lua mutation
• avoid retries, which could double-count an INCR after an ambiguous response timeout

Why

An AOF-backed Redis on a busy host can briefly exceed 500 ms during fsync. The relay currently turns that bounded durability stall into shared rate-limit admission unavailable and can drop an otherwise valid request. Three seconds covers observed sub-1.5-second stalls while keeping a finite failure bound.

Testing

• cargo fmt --check
• cargo test -p buzz-pubsub (26 passed, 11 Redis-gated ignored)
• cargo clippy -p buzz-pubsub --all-targets -- -D warnings
• deployed commit-pinned relay image on Folio
• authenticated query succeeded while Redis was intentionally paused for 1.2 seconds; no admission-unavailable, 503, or relay error
• post-deploy 100-request authenticated read burst passed

[github-actions]

🔐 Codex Security Review

Status: review required for the current range.

The current range is 00e61eafa917d296104006576b7a2ddbfd58bb5a...6e5f47dbbca872032da1c3400c31a76a2e3e3bd7.
A new review must complete for this exact range. When manual authorization
is required, a Block organization member must comment exactly
@buzz-security-review 6e5f47dbbca872032da1c3400c31a76a2e3e3bd7 to authorize a new review.
Any previous review applies only to its recorded range.