Incorrect check for final block not in canonical chain in forkchoiceUpdated(...)

[tersec]

Copying out from #2960 (comment) since it's semi-separate from forked-layers database issue per se and worth addressing.

nimbus-eth1/execution_chain/beacon/api_handler/api_forkchoice.nim

Lines 168 to 171 in a1a9c6b

	if not ben.chain.isCanonical(finalizedBlockHash):
	warn "Final block not in canonical chain",
	hash=finalizedBlockHash.short
	raise invalidForkChoiceState("finalized block not canonical")

is not correct.

It's perfectly valid for there to be two forks, neither of which (yet) finalizes, and for whatever reason, a CL (and thus EL via fcU) has been following one of them, but the other finalizes. At that point, typically the finalized checkpoint (in CL terminology) or from the perspective of the EL/fcU, finalized block, will not necessarily be part of what ForkedChain currently considers to be the canonical chain (which after all, can reorg).