ci: turn the macOS x64 job back on

[WizzardSK]

libretro-build-osx-x64 and its include are commented out. I built that target on a GitHub runner and it produced a core, so this turns the job and the include back on and changes nothing else.

While I was in here: two things about the Android side

Not part of this PR, but worth writing down somewhere.

android-x86 and android-x64 have no libretro OSD path. makefile gives android-arm and android-arm64 each an ifeq ($(OSD), retro) branch — the "RETRO HACK no sdl for libretro android" one that calls GENie with --osd=retro --NO_OPENGL=1. The two x86 targets only have the SDL variant, so a build for them goes down PROJECTDIR_SDL and stops in GENie:

3rdparty/genie/bin/linux/genie ... --osd=sdl --targetos=android --PLATFORM=x64 gmake
Error: invalid option 'with-bundled-sdl2'
make[1]: *** [makefile:1279: build/projects/sdl/hbmame/gmake-android-x64/Makefile] Error 1

That is presumably why android-x86_64 is commented out too. I tried giving those two targets the same retro branch the ARM ones have; it does get past GENie — the target becomes build/projects/retro/hbmame/gmake-android-x64/Makefile — but the build then still stops about thirty seconds in, during the generation phase, and I have not pinned that one down. Happy to send the patch if it is useful as a starting point, but I did not want to propose a fix I cannot show working.

The green android-arm64-v8a job is not building from scratch. In the last pipeline it finishes in 83 seconds, where osx-arm64 takes 1513s and windows-x64 1243s. It does produce an artifact — hbmame_libretro_android.so.zip is in the nightlies — so it is running from the CI cache with the project makefiles already generated, which means the GENie step above never runs. If that cache is ever lost, that job will hit the same wall the x86 ones do today, since REGENIE=0 will no longer have anything to reuse.