fix(pool): Fix Close. Do healthchecks only on idle conns

[ndyakov]

In a rare corner case (actually observed in the CI), if the pool is closed while a connection is in use there will be a data race, since the connection that the pool.Close is trying to execute a healthcheck on (which peeks the socket) is currently being used.

---

Note

Medium Risk
Changes shutdown behavior in the connection pool by skipping socket/reader peeks for non-idle connections, which reduces race risk but could mask close errors for in-use connections during Close().

Overview
Fixes a shutdown-time data race by changing ConnPool.Close() to run isHealthyConn only for connections that are currently idle, and to treat non-idle connections as healthy for the purpose of deciding whether to surface close errors.

Also updates Conn.readerMu documentation to reflect that the reader lock is used for both handoff and health-check-related peek operations.

^{Reviewed by Cursor Bugbot for commit 91a40cf. Bugbot is set up for automated code reviews on this repo. Configure here.}

[jit-ci]

🛡️ Jit Security Scan Results

✅ No security findings were detected in this PR

---

^{Security scan by Jit}