This is a follow up to #39506, #39507 and #39050 to investigate Tamar's suggestions for improving ARM64 perf in ASCIIUtility and Utf16Utility. The suggestions are #39506 (comment), #39507 (comment) and #39507 (comment)

cc @carlossanlop @jeffhandley @kunalspathak @echesakovMSFT @tannergooding