16-bit (ushort) diagonal calculation

[Turnerj]

Adds support for a ushort fast-path in the diagonal calculation, allowing up to 16x items to be processed at once.

Its approximately 15% faster for strings smaller than ushort.MaxValue in a 8000 character benchmark.

Int32

Method	Mean	Error	StdDev	Gen 0	Gen 1	Gen 2	Allocated
Quickenshtein	26.68 ms	0.213 ms	0.189 ms	-	-	-	-

ushort

Method	Mean	Error	StdDev	Gen 0	Gen 1	Gen 2	Allocated
Quickenshtein	23.15 ms	0.112 ms	0.105 ms	-	-	-	-