Under high load of sessions, the connection pool is allowing new
connections to be created.
This is happening because, while the development of `AuthTokenManager`
and `connection liveness check`,
the methods do validate connection on acquired and return to the pool
need to be async.
This changes creates a situation of racing condition which doesn't exist
the original code and this introduces this bug.
Increasing the resouce acquired count before validating the connection
solves the issue, since the race condition is removed.
However, the pool needs also to release the resource when the validation
return `false` or fails for some reason.
This is important to avoid broken connection still be count for the max
pool size.
Co-Authored-By: Max Gustafsson <max.gustafsson@neo4j.com>

The test were to tight and any slowness might cause the test to fail.
Increase the time waiting for the connection get release to pool reduces
the likehood of this problem happens.