insertGlobal hashes inputs twice

[Bodigrim]

insertGlobal calls calculateHash and invokes lookup, which calculateHash again. Hashing could be relatively expensive for large inputs, it would be great to avoid doing it twice.

haskell-symbolize/src/Symbolize/SymbolTable.hs

Lines 80 to 94 in fa328e8

	insertGlobal ba# = do
	GlobalSymbolTable gsymtab sipkey <- globalSymbolTable
	let !key = calculateHash sipkey ba#
	-- SAFETY: If the table IORef contested,
	-- this might trigger `weak` creation for the same bytestring from multiple threads
	-- at the same time.
	-- But finalization is idempotent, and only when a thread finally wins the Compare-and-Swap
	-- will its `weak` pointer be inserted (or alternatively another previously-inserted `ba` returned).
	-- So once this function returns, we can be sure we've returned a deduplicated ByteArray
	!weak <- mkWeakSymbol ba# (removeGlobal key)
	IORef.atomicModifyIORef' gsymtab $ \table ->
	case lookup ba# sipkey table of
	Just ba -> (table, ba)
	Nothing ->
	(insert key weak table, ByteArray ba#)

haskell-symbolize/src/Symbolize/SymbolTable.hs

Lines 118 to 121 in fa328e8

	lookup ba# sipkey (SymbolTable table) = do
	let !key = calculateHash sipkey ba#
	weaks <- IntMap.lookup (hashToInt key) table
	Foldable.find (\other -> other == ByteArray ba#) (aliveWeaks weaks)

[Qqwy]

In theory since lookup should be inlined, CSE might trigger. But instead of hoping that that happens, I think we can just lift the computation out of lookup and pass it explicitly (and calculate it once in insertGlobal and lookupGlobal).