[abseil] Revert changes about arm build

[JackBoosY]

Correct PR #11827, fix build on arm64-uwp.
Since the arm32 build issue is an upstream bug, revert the modification about it.

Tested arm-windows, arm64-windows, arm-uwp, arm64-uwp, x86-uwp and x64-uwp.

Fixes #13929.

[JackBoosY]

There may still be problems with this official patch. I have opened an issue (google/cctz#173) upstream to discuss this changes.

[BillyONeal]

Considering they don't support x86 either it seems unlikely to me that they will care about supporting arm

[BillyONeal]

I don't think this change is correct: it "fixes" the build by removing the /alternatename but upstream comments indicate the alternatename is necessary for customers. The right fix probably involves authoring an /alternatename with the right name mangling for arm64.

[BillyONeal]

I/we should try to update absl to see if they already have these patches fixed, then I think we should outright remove those associated patches.

[JackBoosY]

@BillyONeal Trying...

[JackBoosY]

No, the latest also breaks 32 bit arm build.

[BillyONeal]

@JackBoosY It breaks arm32 but my argument is that the arm32 build was never correct. It builds successfully only because our old patch is breaking this "alternatename" bit -- not emitting an "alternatename" at all. So the linker doesn't look for the nonexistent symbol so the build succeeds, but the resulting library, which presumably depends on that "alternatename" for users is broken. (And likely nobody noticed because we have relatively few customers who would try to use the resulting library on ARM)

I think we should avoid trying to patch in features like this when it's an upstream lack of support that depends on such a fragile patch.

[JackBoosY]

I opened an upstream issue: abseil/abseil-cpp#829

[JackBoosY]

When merging this PR, #11820 should be opened again.

[JackBoosY]

@BillyONeal Done.

[BillyONeal]

Thanks :D