[llvm] [aarch64][win] Add support for import call optimization (equivalent to MSVC /d2ImportCallOptimization) (PR #121516)

[Daniel Paoliello via llvm-commits]

dpaoliello wrote:

Ok, switched this over to using section number and offset assembly directives, implemented using new `MCTargetExpr` derived types (which avoided the need to introduce any new fixup kinds and localizes the changes to just the COFF streamer and object writer).

This also meant that writing the `.impcall` section is now completely contained in the asm printer, avoiding the need for `.impcall` directives (and the weird failure cases associated with them) and will make the x86 implementation much easier.

https://github.com/llvm/llvm-project/pull/121516