`<xutility>`: Consider enabling unsized `memchr` optimization after solving it for ARM64

[AlexGuteniev]

DevCom-1615707 is Fixed - Pending Release.
This means we can remove this workaround:

STL/stl/inc/xutility

Line 4882 in ef62d3f

if constexpr (sizeof(_Iter_value_t<_It>) == 1 && _Is_sized /* TRANSITION, DevCom-1615707 */) {

This should allow optimizing unsized ranges search when vector algorithms are not enabled. This may be helpful for non-x86.

But the other ARM-specific bug should be taken care of - see https://github.com/microsoft/STL/pull/2434/files#r839094547

[CaseyCarter]

Clarifying: "memchr inspects bytes after the match" will be fixed on x86/x64 in 17.4p2 when the redist is unlocked and we ship a new VCRuntime (VSO-1452447). It hasn't yet been fixed for ARM/ARM64/ARM64EC (VSO-1538014).

"memchr returns NULL when count == SIZE_MAX on arm64" (VSO-1488938) was closed as unreproducible on May 9. Can we still repro that bug?

[CaseyCarter]

A fix for VSO-1538014 was just checked in. I think it will ship in v17.8p1, at which time this issue will become unblocked.

[StephanTLavavej]

Workaround removed by #4083, thanks Alex Guteniev (@AlexGuteniev) for the programmer-archaeology.