Remove unnecessary MEMORY64 block

[sbc100]

The only reason this might be useful is of the EM_ASM block happens to return a bigint, which, in general, should not be that case.

[kripken]

Why can't an EM_ASM return a bigint?

[sbc100]

Why can't an EM_ASM return a bigint?

I mean it could... but it could return a bigint without MEMORY64 too. I guess that question really is, why was this block added in the first place? I think that answer is that it used to be that we treated pointers as bigint under wasm64, so it make it more likely. Nowadays we turn pointers into int53 numbers at the boundary so it no more likely that an EM_ASM block would return a bigint under wasm64 or wasm32.

[kripken]

I see. Makes sense. Lgtm if we have a test of returning a pointer from an EM_ASM that works in MEMORY64 mode.

[sbc100]

I see. Makes sense. Lgtm if we have a test of returning a pointer from an EM_ASM that works in MEMORY64 mode.

We actually have a separate EM_ASM_PTR and emscripten_asm_const_ptr to the times when you want to return a pointer.

[kripken]

Nice, I didn't know we had EM_ASM_PTR.

[sbc100]

I think I'm going to land #18218 first which refactors stuff here.