Feature Request

Summary

We re-compute the block hash several times during consensus. We should just cache this block hash computation. E.g. each HashesTo call recreates the block's hash.

Here is the time spent in block hashing in a node synced to osmosis mainnet for 1 hour:

This is by no means the bottleneck, but is part of the sequential bottleneck of consensus and would be good to reduce, via only computing it once per block.

Proposal

I suggest that we cache this by adding a verifiedHash field to the internal Block type.