app: drop stateful mempool

[erwanor]

Describe your changes

This is a draft PR that removes the stateful mempool from the application, replacing it with:

• a stateless mempool that gates inclusion on validity against the latest version of the app state
• a prepare proposal implementation that filters for valid transactions
• a process proposal implementation that validates block quality and lays the groundwork to do optimistic execution

Checklist before requesting a review

• If this code contains consensus-breaking changes, I have added the "consensus-breaking" label. Otherwise, I declare my belief that there are not consensus-breaking changes, for the following reason:

  Consensus breaking

[conorsch]

Really solid work here. Especially grateful for the careful commenting throughout, as this work builds on a lot of recent and rather subtle refinements to the constraints on message sizes acceptable to the app.

[conorsch]

Not merging because review from @hdevalence is still outstanding. Once it's got final approval, @erwanor, let's use the new squash-with-commit-comments-preserved workflow we've been discussing to get it in!

[hdevalence]

Did we ever choose these in a principled way? Should they be parameters rather than hardcoded constants?

[erwanor]

This was picked with three things in mind:

• large enough to allow JIT liquidity for multiple pairs
• large enough to allow a ranged collection of positions
• large enough to allow multiple IBC client updates
• small enough to allow many of those actions to be performed in a single block even in the extreme case where each transaction is one of those things

Making them parameters make sense, I didn't do it because it would require more work with penumbra-governance and hooking them into https://docs.rs/tendermint/latest/tendermint/consensus/params/struct.Params.html

[hdevalence]

I think we should consider making the constants be parameters, but I'm good to merge either way.