[CIR][CIRGen] Handle Address Space Casting on builtin allocas

[RiverDave]

Replaces the assertion with a call to performAddrSpaceCast for __builtin_alloca__

• Converts the type and passes the appropriate address space attributes
  Added a new test file clang/test/CIR/CodeGen/dynamic-alloca-with-address-space.c that verifies:
  • CIR generation for dynamic allocas with address spaces
  • LLVM IR generation from CIR matches expected output
  • Generated LLVM IR matches the output from the original code generator

The test specifically targets the AMD GPU target (amdgcn-amd-amdhsa) which uses non-default address spaces.

[RiverDave]

• [CIR][CIRGen] Handle Address Space Casting on builtin allocas #2064 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.