In MEMORY64 mode, function pointers are 64-bit, but would be better to be 32-bit

Currently it looks like when compiling to -sMEMORY64, function pointers are 64-bit integers, and as result, e.g. {{{ makeDynCall() }}} macro does not currently work.

I cannot think of a technical reason why we would want to have 64-bit function pointers, since these will always be count-from-zero indexes to the wasm function table anyways. Would it be possible to make LLVM treat function pointers as 32-bit integers?

(soft reasons may be wanting to keep function pointers and integer pointers sizes consistent with each other, or to simplify e.g. some LLVM implementation)