Note

Medium Risk
Changes reconnect behavior under slow or stuck Redis/network by capping ping and reconnect duration; mis-tuned timeouts could trigger reconnects sooner than before, but defaults are reasonable for a background health loop.

Overview
Fixes #3811 by making the background PubSub channel health-check goroutine use bounded deadlines instead of an unbounded context.TODO() for periodic Ping and for reconnect after a failed ping.

Adds WithChannelPingTimeout (default 5s) and WithChannelReconnectTimeout (default 10s) as ChannelOptions, stored on the internal channel struct. When the health-check timer fires, Ping runs under context.WithTimeout(..., pingTimeout); on failure, reconnect runs under a separate timeout (default 10s).

Also adjusts timer draining when a ping is received on the c.ping channel: it uses a non-blocking drain of timer.C instead of timer.Stop() plus blocking read, avoiding a possible block on the timer channel.

^{Reviewed by Cursor Bugbot for commit 9719406. Bugbot is set up for automated code reviews on this repo. Configure here.}