storage: prototype and evaluate expected inflation of Raft proposal sizes

[tbg]

See #6166. We want to get a good idea of the proposal sizes which typical (and perhaps pathological) workloads generate. This size is (for all practical purposes) dictated by the size of the generated WriteBatch representation.

[tbg]

No real surprises here. I looked at batch puts, and surprise surprise the repr size scales like (keyBytes+valBytes). More concretely, a raw payload of 10 value bytes and an 8-byte key wound up about double that in the repr (but that includes MVCC overhead, etc). 100000 * 18kb gave a ~352KiB repr.

DeleteRange is similarly without surprises. The nice thing here is that the deleted values don't matter - it's always O(encoded keySize).

All of this is obvious from the encoding, so this was really just a sanity check. If there's anything else worth looking at, ping this issue. Numbers are from the (hacky) fiddle-6285 branch.