This commit changes the way compilation is done using CMake. As its a mess in CMake under Apple target as its actively injects parameters Arch, SYSROOT parameters.
That makes compilation for different arches for iOS/MacOS cross compilation problematic as we can't control these params (already present from CMake).
The only way to make it do what expected is to tell CMake we are doing cross compilation by specifying `CMAKE_SYSTEM_NAME`.
In this case it will try to provide SYSROOT but it can be controlled by overriding `CMAKE_OSX_SYSROOT`.

Sources for this lib was missing in project.
Picked from https://github.com/robovm/hfscompressor, adapted to build using CMake and recent dependencies.
Added option to compile for both `macosx-arm64` and `macosx-x86_64`.
`HfsCompressor.java` update to load platform specific binary

This commit adds `Environment (Native/Simulator)` option that allows makes targets different when building for Arch.
For example build for `arm64` now could be targeted to `ios-arm64`, `macosx-arm64`, `ios-arm64-simulator`.
LLVM detects proper target from triple and it is being generated now with respect to `Environment` option.
Also linker `-miphoneos-version-min=` parameter was replaced with `--target=` one that accepts the tripple.
Changes were done to folder names: environment is being attached to every arch name where applicable. For example:
```
vm/lib/arm64/
vm/lib/arm64-simulator
vm/lib/thumbv7
vm/lib/x86-simulator
vm/lib/x86_64-simulator
```

CMake was adapted to workaround mess on Apple platform. Using cross compilation mode.
Targets were extended to produce binaries for additional targets: `macosx-arm64`, `ios-arm64-simulator`

Logic addapted to following:
- simulator considered to have `arm64` arch only on m1 host and if its ios version is 14+
- x86 simulator is not available on `arm64` m1 host

- console target now has arch option that allows to select x86_64/arm64 target on m1 silicon;
- `auto` simulator arch is set to the on of the host. E.g. on m1 will be set to `arm64`

These issues were discovered during run against debug version of llvm lib and crashes happened on assertions:
* llvm: changed Load/Store atomic access ordering from `unordered` to `monotonic` as it was crashing during AtomicExpandPass for x86 target. Line of code where it crashing `getStrongestFailureOrdering`. In release build with assertions removed it falls down into `Monotonic` case anyway.
* bug fixed: wrong calculation of struct offset for Vectorized Structs. Root case is that vector was wrapped into single element struct like `{<2 x float>}` and any try to get offset for member 1+ returns garbage (or crash on assert in debug version of llvm). Workaround -- getting element storage size and multiply by index

while `macos-latest` is 10.15 its not able to build m1 `macosx-arm64` target. But `x86_64` can be build.
these allows to build libhfs, libmobiledevice, libllvm natives. In case of success build new branch with artifacts is pushed.
It can be then merged into master.
Build are triggered manually and are parametrized:
- "Arguments to build.sh" -- allows specify single target to build;
- "Target branch to push artifacts" -- name of the branch were built artifacts will be pushed;
- "Commit message" -- message to be used while committing changes

…target` with triple is used for this now

…older hw

…nitial_support

…ial_support