1. Fixes a tricky case when compiling on AppleSilicon (arm64) for x86_64 (happens, e.g., when cmake is x86_64).

All libraries are built for x86_64, but htslib checks compile for arm64 and fail linking against zlib compiled for x86_64.
Now we explicitly pass the target architecture for htslib.

2. Also, now we always link against the static locally compiled htslib binary.

3. Also improvements in the CI:

• compile with more threads
• correctly set the platform (previously all were Apple Silicon)
• check x86_64 build on arm64