Re-enable UseAggressiveReg2MemForCodeSize on Wasm

[kateinoigakukun]

Description

The optimization is currently disabled on Wasm as it causes LLVM ISel panic.

PromoteFloatResult #0: t51: f16 = freeze t7

<unknown>:0: error: fatal error encountered during compilation; please submit a bug report (https://swift.org/contributing/#reporting-bugs)
<unknown>:0: note: Do not know how to promote this operator's result!

Full failed build log

Once we fix this issue, we should revert the following changes:

• Disable UseAggressiveReg2MemForCodeSize on WASM #77502
• Disable two tests on wasm #77509
• Try harder than #77509 to disable tests on wasm #77518

[kateinoigakukun]

As far as I investigated, it looks like an LLVM-side issue and other architectures have similar issues:

• [loongarch][DAG][FREEZE] Fix crash when FREEZE a half(f16) type on loongarch llvm/llvm-project#107791
• LLVM miscompiles passing/returning half on several backends by using lossy conversions llvm/llvm-project#97981
• [MIPS] Use softPromoteHalf legalization for fp16 rather than PromoteFloat llvm/llvm-project#110199

[kateinoigakukun]

Here is a minimal crash repro:

target triple = "wasm32-unknown-unknown"

define half @freeze_half(half %f) nounwind {
  %t0 = freeze half %f
  ret half %t0
}