Use math/rand.Rand.Read to generate slices of random data

[mtrmac]

… instead of open-coding loops, primarily to simplify.

In one case this is a speedup, in the other it’s a noticeable slowdown, a correctness improvement in theory at least.

[mtrmac]

(If you prefer the code as is, feel completely free to close this PR.)

[vrothberg]

LGTM
@giuseppe @nalind @rhatdan PTAL

[mtrmac]

FWIW

% go test -v -bench . -benchmem ./bench_test.go
goos: darwin
goarch: amd64
cpu: Intel(R) Core(TM) i7-9750H CPU @ 2.60GHz
BenchmarkStandard
BenchmarkStandard-12       	      28	  38833565 ns/op	52434254 B/op	       2 allocs/op
BenchmarkSourceOnly
BenchmarkSourceOnly-12     	      22	  51455058 ns/op	52428800 B/op	       1 allocs/op
BenchmarkSourceOnly8
BenchmarkSourceOnly8-12    	      22	  50682468 ns/op	52428903 B/op	       1 allocs/op
BenchmarkUnsafeSlice
BenchmarkUnsafeSlice-12    	      56	  20757283 ns/op	52428803 B/op	       1 allocs/op

on 50 MiB output, with

• Standard: rand.New(rand.NewSource(seed)).Read
• SourceOnly: Two nested loops turning Int63 into 7 bytes, like one of the previous implementations
• SourceOnly8: Two nested loops turning Int63 into 8 bytes (it turns out the compiler doesn’t unroll that any better)
• UnsafeSlice: A single loop storing Int63 into an int64, like the other previous implementation

[giuseppe]

LGTM