Backport: Core: Add pipeline send timeout and fail pending requests during recovery (#5755)

[jduo]

Summary

Backport of #5755 to release-2.2. Adds a 100ms timeout on pipeline.send() to detect dead connections (half-open TCP) and fails pending requests immediately when entering recovery.

Issue link

Closes #5715
Closes #5716

Features / Behaviour Changes

• pipeline.send() now times out after 100ms with FatalSendError instead of blocking forever when the bounded channel (50 slots) is full due to a dead connection
• poll_recover() properly polls JoinHandle instead of using now_or_never(), fixing waker registration
• poll_flush() returns Pending during recovery instead of busy-spinning in a loop{}
• New fail_pending_requests() drains pending request queue with ClientError("Connection in recovery") when recovery is in progress
• Reconnection operations are non-blocking (tokio::spawn)

Limitations

Without this fix, a single dead shard blocks the pipeline indefinitely, preventing recovery from triggering and causing OOM under sustained load.

Testing

Validated on EC2 against ElastiCache (3-shard CRR cluster):

• iptables block on one primary for 30s → client recovers within 5s of block being lifted
• Previously (without fix): client stuck indefinitely, never recovered

cargo check, cargo clippy, cargo fmt all pass.

Checklist

Before submitting the PR make sure the following are checked:

• This Pull Request is related to one issue.
• Commit message has a detailed description of what changed and why.
• Tests are added or updated.
• CHANGELOG.md and documentation files are updated.
• Linters have been run (make *-lint targets) and Prettier has been run (make prettier-fix).
• Destination branch is correct - main or release
• Create merge commit if merging release branch into main, squash otherwise.

[alexr-bq]

Approved pending fixing CI