Fix freeze-crash in lightmapper under MinGW-GCC (3.2)

[RandomShaper]

Why not another attempt at fixing #45097?

I've found a non-hacky way to emit the PAUSE instruction, via an intrinsic that seems to work. Something feels different... The Preparing data structures step now is almost immediate.

I've tested with one more project this time.

@akien-mga, a couple of remarks:

• I'm not actually building on Fedora, but on Windows with MinGW-w64. I know the results may vary, but they seem to be very similar most of the time.
• I've "reconciled" the new patch with the stored one manually, so the patch file may be wrong. (Reverse-applying the patch in order to get a new one with all the changes didn't work for me; I don't know why.) I guess if this works we can take care of that later.

[akien-mga]

You're amazing :D
I'll throw a build with that PR to get another round of testing from our trustworthy crash-testers :)

[akien-mga]

Thanks!

[akien-mga]

I'm not actually building on Fedora, but on Windows with MinGW-w64. I know the results may vary, but they seem to be very similar most of the time.

For the reference, what MinGW and GCC versions do you have? You might have newer ones than I have on Fedora (MinGW 7.0.0 and GCC 10.2) which might have solved some of the alignment issues reported upstream?

[RandomShaper]

@akien, build-info.txt