[mono][aot] optimize aot Unsafe.BitCast for more cases

[lewing]

Exrtacted from #115062

VType bitcast should work fine on wasm,

digging through the history some related prs (this will not address all of them)
#105150
#103915
#104694
#106430 (not fixed by this pr)

[Copilot]

Pull Request Overview

This PR enables xcast support for wasm by updating type conversions and removing target size restrictions for SIMD operations. Key changes include:

• Introducing conditional type mapping for size 4 and size 8 conversions.
• Removing the TARGET_SIZEOF_VOID_P check for SIMD xcast operations to allow them unconditionally.

Comments suppressed due to low confidence (1)

src/mono/mono/mini/intrinsics.c:757

• Consider verifying that removing the TARGET_SIZEOF_VOID_P check for SIMD xcast is intentional, as this may inadvertently enable xcast on non-wasm platforms.

opcode = OP_XCAST;

[lewing]

I don't know if we are running any aot tests on non-wasm 32 bit platforms at this point? I can easily limit it to Wasm. There may have been a reason @fanyang-mono excluded it initially in #105150

[steveisok]

I don't know if we are running any aot tests on non-wasm 32 bit platforms at this point? I can easily limit it to Wasm. There may have been a reason @fanyang-mono excluded it initially in #105150

I think it's ok limiting it to Wasm. The majority of our tests on arm android is JIT based.

[lewing]

/ba-g timeouts are not related