Checkpoint blocks that cannot be processed properly

[geoff-vball]

Context and scope
Currently in listener.go, the select statement checking for new blocks can fail at multiple points and we will just continue and skip processing if there is a networking error or otherwise. This will currently cause us to never checkpoint that block, which means the next time the relayer restarts, it will try to reprocess from that block, which may cause a lot of unnecessary reprocessing.

Any error should probably still checkpoint the block, but we should save that block to attempt to be reprocessed somehow. We can potentially save that block number to the database to attempt to be reprocessed periodically.

Discussion and alternatives
This method of reprocessing can probably be extended in the case where we chose not to process a message because it is not currently profitable, but may be profitable in the future.

Open questions