Add connection pooling to Trino Gateway backend datastore

[Raman Sarvaria]

Out-of-the-box JDBI does not seem to leverage connection pools and will create a new connection for every query against the backend datastore as seen in the public Slack channel here. During high traffic this can lead to reaching maximum connection counts against the backend datastore.

It has been noted that there are multiple ways of getting around the limitation of maximum number of connections:

1. Optimize the underlying datastore configurations so that connections drop off quicker
2. Use pgbouncer as a service in between Trino Gateway and PostgreSQL (assuming postgres is used as the backend store instead of MySQL or similar)
3. Implement connection pooling as part of the Trino Gateway code itself. Example PostgresSQL connection pooling provided by JDBI docs: https://jdbi.org/#_high_availability

As an enhancement we are looking to minimize overall connections required by Trino Gateway against its backend datastore.

[Manfred Moser]

Trino already has connection pooling in a number of places. Not sure if they involved JDBI though ... we definitely have some for JDBC connections though. We can probably take some inspiration from there. Or maybe there even is something in airlift (although I dont remember seeing something along that line).