[Watcher] Retries considerations

[greged93]

Raising this issue linked to a bug that was found when using the l1geth endpoint as l1.url. The provider never finds transaction with hash 0xa05084ac5b23759f89069dbf64c390a64903afd5475c5f053f00808cc8126f37 which is a commit batch on Sepolia (https://sepolia.etherscan.io/tx/0xa05084ac5b23759f89069dbf64c390a64903afd5475c5f053f00808cc8126f37). This causes the watcher to error and go in a endless loop, because it can't advance, reindexing previous events.

We should consider what to do after the watcher fails to advance for a defined amount of retries.