Add logic to restart pool on checkout_timeout error.

[rkachowski]

Description

There has been a spike in checkout_timeout errors since the mainnet stall - https://cloudlogging.app.goo.gl/Jnk5oDzFiee9nrqYA . According to edgurgel/httpoison#414 it seems that an entire http connection pool can be affected by timeout errors when the pool size is suitably large.

This PR includes the suggested fix to stop the hackney connection pool when this error is encountered.

Testing

• ran locally against staging archive node and asserted that http connection to archive nodes was running successfully

Unclear how to test a situation where checkout_timeout is triggered directly, but we will rollback if affected

[github-actions]

Unit Test Results

       4 files  ±0     326 suites  ±0   2m 58s ⏱️ -7s
2 250 tests ±0  2 205 ✔️ ±0  45 💤 ±0  0 ❌ ±0 
2 281 runs  ±0  2 236 ✔️ ±0  45 💤 ±0  0 ❌ ±0 

Results for commit 9620610. ± Comparison against base commit b298c91.

♻️ This comment has been updated with latest results.