fix(pool): prevent infinite loop during idleTimeout cleanup in transaction

[thanathip41]

Problem:

• When a transaction remains open longer than the configured idleTimeout, the pool may destroy the underlying connection.
• This results in an Error: idle timeout being thrown while the transaction is still active.
• In some cases, the error repeatedly propagates through the cleanup logic, causing an infinite loop that continuously emits errors and consumes resources.
• The process may become unresponsive or eventually crash.

Root Cause:

• The idle timeout cleanup path can repeatedly trigger error handling for the same connection state.
• Instead of terminating cleanly, the process may enter a continuous error loop.

Fix:

• Detect the idle timeout failure condition and terminate the process immediately.
• Avoid repeatedly executing the cleanup/error handling path for the same connection.
• Fail fast rather than allowing the process to enter an infinite error loop.

Test:

• Added a test covering a long-running transaction that exceeds the configured idleTimeout.
• Verified that the process exits cleanly instead of repeatedly throwing Error: idle timeout.
• Verified that normal idle connection cleanup behavior remains unchanged.