Optimizing the following String APIs

1. String.Split --> Optimizing MakeSeparatorListVectorized
2. String.Replace(char oldChar, char newChar) --> Optimizing for a single iteration. Although we have measured perf on this API, it just represents optimizing a single iteration and not all.

PERF on ICX

Below tables show a result comparison output by ResultComparer in the performance repo.

Base = No changes
Diff = With the PR changes

1. Split

A Vector128 code path already exists for this API. We are adding a similar Vector256 and Vector512 code path.

base =Diff Vector256 code path vs diff =Diff Vector512 code path

base =Base Vector128 code path vs diff =Diff Vector256 code path

This is one of the issues where Avx512 is not that performance because of the issue with using multipleVector512.Equals(). I ran a couple of iterations using StopWatch method and below are the results.

As you can see, for each iteration,Vector512 is almost the same asVector256. Let me know if there are any suggestions for further optimizingVector512 code path. We have to decide whether this can ne merged or not since there are alreadyVector128 code path for both the APIs. Also, the Vector256 and Vector512 code path provide a significant speed up over Vector128 code path.

2. Replace_Char

A Vector128 code path already exists for this API. We are just adding a single iteration of Vector512 or Vector256.

base =Diff, Vector256 code path vs diff =Diff Vector512 code path

base =Base Vector128 code path vs diff =Diff Vector256 code path