Messagepool lower priority optimizations

[snissn]

What is the motivation behind this feature request? Is your feature request related to a problem? Please describe.

related to issue #10561 perf: message pool: change locks to RWMutexes for performance #10561 are the following issues that were identified that are to be added in a future pull request. The motivation for the two separate PRs is to isolate the threading changes as much as possible for correctness reasons.

• 1. https://github.com/filecoin-project/lotus/pull/10561/files#r1148322934 in chain/messagepool/check.go pre-allocate the size of the array msgs array using msgs = make([]*types.Message, 0, len(mset.msgs))
• 2. https://github.com/filecoin-project/lotus/pull/10561/files#r1152024072 remove the requirement of mp.keyCache.Add(ka, ka) by checking that the address is a robust address
• 3. https://github.com/filecoin-project/lotus/pull/10561/files#r1150875872 chain/messagepool/repub.go - optimize the Lock timing and confirm if all tests pass with a read lock -- also ensure that mp.curTsLk.RLock() is not held during mp.api.ChainComputeBaseFee
• 4. https://github.com/filecoin-project/lotus/pull/10561/files#r1150137317 chain/messagepool/selection.go change to read lock
• 5. in chain/messagepool/check.go pending has a giant warning saying do NOT use it directly -- we should probably be calling getPendingMset here for address resolution?

moving the sixth item to its own issue for further discussion:

#10692