Improve pindexer speed

[cronokirby]

It takes too long to sync up pindexer on a raw database: on the order of hours. The processing logic we do is not so complicated as to make this take this long, and I suspect that the architecture of processing one event at a time and touching the database for every event makes this slow.

I think we should move to have a "per batch of blocks" processing logic, only reading the database before the batch, and writing after the batch.

This also has the advantage of making certain app views simpler, because right now they would like to process all the events in a block at once, e.g. to know the timestamp for each event, but have to resort to tricks like using an in database queue instead.

[conorsch]

It takes too long to sync up pindexer on a raw database: on the order of hours

As of #4913 & #4917, running pindexer against a db for the penumbra-1 chain takes ~25m, assuming block height of 2.2 million. That's a huge performance increase, enough that I think we can close this issue. Unless there are other enhancements you've in mind @cronokirby that are worth recording here?

[cronokirby]

Awesome! Yeah we can close this for now, the low hanging fruit mentioned in this issue have been resolved.