fix MSM bench using 64-bit scalars after #444 [skip ci]

[mratsim]

While implementing #444, I changed the MSM benchmarks to use 64 bits scalars (and 128 and 192 ...) to evaluate the endomorphism bit-threshold

constantine/constantine/named/zoo_endomorphisms.nim

Lines 110 to 129 in 618d8b3

	func hasEndomorphismAcceleration*(Name: static Algebra): bool {.compileTime.} =
	Name in {
	Bandersnatch,
	Banderwagon,
	BN254_Nogami,
	BN254_Snarks,
	Secp256k1,
	BLS12_377,
	BLS12_381,
	BW6_761,
	Pallas,
	Vesta
	}
	const EndomorphismThreshold* = 192
	## We use substraction by maximum infinity norm coefficient
	## to split scalars for endomorphisms
	##
	## TODO: explore an alternative way to split scalars, for example via division
	## https://github.com/mratsim/constantine/issues/347

Unfortunately, 64 bits was checked in and performance figures were misleading and up to 3x faster than expected in fef6fdc from 3 weeks ago. This has not made the v0.1.0 release from 7 weeks ago.

zkalc integration #425 is also correct.