fix(indexer): retry write on PG deadlock

[onewland]

Not sure of a clean way to test this, but if people want me to hack around and force a deadlock error I can try to do so

https://getsentry.atlassian.net/browse/SNS-1519

[onewland]

i would add some sort of exponential sleep backoff here, but generally not sure if this is the sanest way to deal with those errors

does the indexer currently skip those messages, or crash entirely?

I haven't personally verified this but according to the ticket the messages get dropped (not a full crash).