[llvm] [AMDGPU] Limit promoting allocas that have users with dynamic index above a threshold on number of elements (PR #170327)

[Kevin Choi via llvm-commits]

choikwa wrote:

It would seem that adding 2x BITCAST per User is still beneficial over scalarized copies, and that work can precede the disabling alloca check which doesn't check for alignment. Will look into that first.

https://github.com/llvm/llvm-project/pull/170327