Better benchmarks for intersection (and better intersectionArraysBy)

[sjakobi]

Here are the intersection benchmarks as of #406:

unordered-containers/benchmarks/Benchmarks.hs

Lines 322 to 325 in d24cc1f

	, bgroup "intersection"
	[ bench "Int" $ whnf (HM.intersection hmi) hmi2
	, bench "ByteString" $ whnf (HM.intersection hmbs) hmbsSubset
	]

unordered-containers/benchmarks/Benchmarks.hs

Lines 84 to 88 in d24cc1f

	elemsBS = zip keysBS [1..n]
	keysBS = UBS.rnd 8 n
	elemsI = zip keysI [1..n]
	keysI = UI.rnd (n+n) n
	elemsI2 = zip [n `div` 2..n + (n `div` 2)] [1..n] -- for union

unordered-containers/benchmarks/Benchmarks.hs

Lines 103 to 107 in d24cc1f

	hmbs = HM.fromList elemsBS
	hmbsSubset = HM.fromList (takeSubset n elemsBS)
	hmi = HM.fromList elemsI
	hmiSubset = HM.fromList (takeSubset n elemsI)
	hmi2 = HM.fromList elemsI2

unordered-containers/benchmarks/Benchmarks.hs

Lines 120 to 123 in d24cc1f

	takeSubset n elements =
	-- use 50% of the elements for a subset check.
	let subsetSize = round (fromIntegral n * 0.5 :: Double) :: Int
	in take subsetSize elements

I think the maps in these benchmarks have an unusually large overlap – there are relatively few elements that do not lie in the intersection. It would be interesting to have benchmarks where the intersection is relatively smaller.

With such benchmarks it may also turn out that it is faster to iterate only over the intersections of the array nodes in intersectionArrayBy, as implemented in oberblastmeister@1ee67ea.