feat(mempool): remove redundant locks

[swift1337]

Removes mutex calls from high-level mempool as underlying mempools are already thread-safe.

Local tests showed that insertEVM took up to ~10ms for the lock, and SelectBy up to ~80ms, which is a lot for consensus progression.

Note: txpool.Add() is still slow because it also has a lock inside geth's part of the codebase and also perform some operations under this lock. Local bench showed that it might take up to ~10ms for this operation,

evm/mempool/txpool/legacypool/legacypool.go

Line 1095 in 48bde58

pool.mu.Lock()

Overall, abci.InsertTx() takes ~30ms

Closes STACK-2019

[linear]

STACK-2019 reduce mutex usage in the mempool