Prevent duplicate code hash computations when tracing `AccountFragment`'s

[OlivierBBB]

Code hashes depend deterministically only on the tuple (address, depNum, depStatus) (with depStatus = true having trivial code hash, anyways). A growing map with such tuples as keys and hashes as values would prevent lots of duplicate code hash computations in

linea-tracer/arithmetization/src/main/java/net/consensys/linea/zktracer/module/hub/fragment/account/AccountFragment.java

Lines 170 to 173 in 49c71be

	.pAccountCodeHashHi(oldState.tracedCodeHash().hi())
	.pAccountCodeHashLo(oldState.tracedCodeHash().lo())
	.pAccountCodeHashHiNew(newState.tracedCodeHash().hi())
	.pAccountCodeHashLoNew(newState.tracedCodeHash().lo())

[letypequividelespoubelles]

For my understanding, this perf is only at commit time, as at sequencer time, https://github.com/Consensys/linea-tracer/pull/2463/files alreday solves the issue, right ?