tracking: performance optimizations

[0xalpharush]

We don't need a state commitment and it's slow so let's mock it and re-execute transactions as needed (we are already doing this for tracing after the latest go-ethereum upgrade). The only trie that is loaded from the db rn is the genesis and/or post-deployment state AFAIK.

medusa/chain/test_chain.go

Lines 828 to 830 in b81a9bc

	// Perform a state commit to obtain the root hash for our block.
	root, err := t.state.Commit(t.pendingBlock.Header.Number.Uint64(), true)
	t.pendingBlock.Header.Root = root

medusa/chain/vendored/apply_transaction.go

Line 64 in b81a9bc

root = statedb.IntermediateRoot(config.IsEIP158(blockNumber)).Bytes()

We should simplify the main fuzz loop to just be changing the block num/timestamp in the block context and not doing all the complex stuff with pending/discarding blocks as if it's a real chain IMO

medusa/chain/test_chain.go

Lines 742 to 743 in b81a9bc

	// Create a new context to be used in the EVM environment
	blockContext := newTestChainBlockContext(t, t.pendingBlock.Header)

[bsamuels453]

consider removing bloom filter calculations (CreateBloom) as well

eats up about 4-5% of the cpu time of EVMApplyTransaction.

if we do this, plan on adding a way to query the list of logs generated by a tx/block since we might still need it for other features

[0xalpharush]

Looking at the CPU report of pprof the most notable improvement we can make is probably memoizing this

medusa/fuzzing/coverage/coverage_tracer.go

Lines 164 to 168 in 3a9b0fa

	// Obtain our contract coverage map lookup hash.
	if callFrameState.lookupHash == nil {
	lookupHash := getContractCoverageMapHash(code, callFrameState.create)
	callFrameState.lookupHash = &lookupHash
	}

EDIT: This is done #472