Fixes pool poisoning

[nihilSup]

Fixes #550

Implemented approach with cleaning up zombie connections before creating new connection for both sync and async implementations. Added new status is_connecting to AsyncConnectionInterface and ConnectionInterface. Added tests for pool poisoning and new code.

Other possible solutions:

• Request errors on overloaded systems may poison the connection pool #550 (comment) prevents stale connections from happening, not sure that it is safe to reuse connection after PoolTimeout in common
• Request errors on overloaded systems may poison the connection pool #550 (comment) in place clean up, need review

[lovelydinosaur]

Thanks so much for taking a look at this. 🙏🏼
Apologies for the delay, I am getting around to reviewing this now.

See also #681

[lovelydinosaur]

Going to close this off since I think it's superseeded by #681 and #719.