Follow-up work on shared allocation helpers

[filipnavara]

PR #115339 enables sharing of the allocation helpers written in assembly between CoreCLR and NativeAOT runtimes. This is a tracking issue for follow-up work that came up during review of the PR:

• Change managed signatures of new array helpers to take nint (1)
• Optimization: Enable the RhpNewObjectArrayFast helper on NativeAOT and teach JIT interface to use it.
• Cleanup: Extract the 0x40000000 magic constant in RhpNewObjectArrayFast to a named constant like FAST_OBJECT_ARRAY_ALLOCATION_THRESHOLD (1)
• Correctness: NativeAOT doesn't properly handle multi-dimensional arrays larger than 2 GB (1, 2, 3)
• Optimization: Investigate and benchmark potential improvements to x86 allocation helpers (1)

[dotnet-policy-service]

Tagging subscribers to this area: @agocke, @MichalStrehovsky, @jkotas
See info in area-owners.md if you want to be subscribed.

[kant2002]

1. According to Fix illumos compile error in coreclr/runtime/amd64/AllocFast.S #117463 the constant 0x40000000 would not appear in source code, but mostly in comments. Is there still need for creation of FAST_OBJECT_ARRAY_ALLOCATION_THRESHOLD if it would be in comments?
2. RhpNewObjectArrayFast is now RhpNewPtrArrayFast. That's for passerbuys.

[am11]

1. According to Fix illumos compile error in coreclr/runtime/amd64/AllocFast.S #117463 the constant 0x40000000 would not appear in source code, but mostly in comments. Is there still need for creation of FAST_OBJECT_ARRAY_ALLOCATION_THRESHOLD if it would be in comments?

That's only for amd64, other assemblies have the constant in use:

src/coreclr/runtime/arm/AllocFast.S:        mov         r2, #1 << 28 // 0x40000000 / sizeof(void*)
src/coreclr/runtime/arm64/AllocFast.S:        mov         x2, #(0x40000000 / 8) // sizeof(void*)
src/coreclr/runtime/i386/AllocFast.S:        cmp         edx, (0x40000000 / 4) // sizeof(void*)
src/coreclr/runtime/loongarch64/AllocFast.S:        li.w  $a2, (0x40000000 / 8) // sizeof(void*)
src/coreclr/runtime/riscv64/AllocFast.S:        li          t0, (0x40000000 / 8) // sizeof(void*)