add retries to avoid relatively benign WARN logs

[arajasek]

Problem: there are 2 aspects of the system that can generate WARN logs a little too zealously. These are the Xchain Provider's FetchBatch method, which may fail due to being very slightly behind another node, and the relayer's send method, which might have submissions be reverted on very dynamic chains due to out of gas (eg. here). In both cases, the components upstream will retry anyway, and likely succeed, but still emit a failure log / metric.

Solution: given that this is relatively benign, let's retry a few times within these methods before erroring.

[arajasek]

I'm not sure I like doing this for the relayer's Send method. We should probably capture that failure immediately.