Add unicode_loose_xxhash Vindex type

[aquarapid]

• Refactor out unicode normalization functions
• Add md5 benchmark
• Change existing xxhash benchmark to benchmark Vindex
  function instead of underlying hash

Addresses #6457

Signed-off-by: Jacques Grove aquarapid@gmail.com

[aquarapid]

For reference here are the benchmark runs:

BenchmarkMD5Hash/md5Hash,direct,bytes,n=8B-32            6178754               187 ns/op          42.71 MB/s
BenchmarkMD5Hash/md5Hash,direct,bytes,n=32B-32           6169424               191 ns/op         167.59 MB/s
BenchmarkMD5Hash/md5Hash,direct,bytes,n=64B-32           4098876               290 ns/op         221.07 MB/s
BenchmarkMD5Hash/md5Hash,direct,bytes,n=512B-32          1159005              1036 ns/op         494.44 MB/s
BenchmarkMD5Hash/md5Hash,direct,bytes,n=1KB-32            667616              1784 ns/op         560.67 MB/s
BenchmarkMD5Hash/md5Hash,direct,bytes,n=4KB-32            166500              6782 ns/op         589.81 MB/s
BenchmarkNormalizeSafe-32                                 141795              9848 ns/op
BenchmarkNormalizeShared-32                              2586418               425 ns/op
BenchmarkNormalizePooled-32                              2234502               451 ns/op
BenchmarkXXHash/xxHash,direct,bytes,n=8B-32             27336464                39.2 ns/op       203.93 MB/s
BenchmarkXXHash/xxHash,direct,bytes,n=32B-32            24836524                47.1 ns/op       679.41 MB/s
BenchmarkXXHash/xxHash,direct,bytes,n=64B-32            23306372                50.2 ns/op      1274.56 MB/s
BenchmarkXXHash/xxHash,direct,bytes,n=512B-32           12211312                96.8 ns/op      5289.02 MB/s
BenchmarkXXHash/xxHash,direct,bytes,n=1KB-32             7712202               148 ns/op        6761.02 MB/s
BenchmarkXXHash/xxHash,direct,bytes,n=4KB-32             2593194               462 ns/op        8657.97 MB/s

[harshit-gangal]

Looks good to me.
We would need an end to end test to verify it works.