Signed-off-by: Rui Chen <rui@chenrui.dev>

 + rename define from LIBDIVIDE_X64_INTRINSICS to LIBDIVIDE_MULH_INTRINSICS
 + __umulh/__mulh available for x64/arm64

__builtin_clz/__builtin_clzll for ms compiler taken from https://gist.github.com/pps83/3210a2f980fd02bb2ba2e5a1fc4a2ef0

Fix Assembler Warning: mnemonic suffix used with `div'

Due to multiple threads running tests, initial message appears to be mixed up with output from other threads

AArch32 also has NEON, but lacks some of the intrinsics needed.
Explicitly check for presence of an AArch64-only intrinsic to
make the NEON check fail on AArch32.

Fixes		#102

* Preserve order of params when calling libdivide_mullhi

This change improves code generation in 32-bit builds (shorter code, uses less stack).

* comments

All these function are declared with `static LIBDIVIDE_INLINE`, but do not have them at definition

Simply check for the value of the `__STDC_HOSTED__` macro before including `<stdlib.h>` or `<stdio.h>` and others.
Moreover, as an idea for future changes, you could also add some checks for user-provided `LIBDIVIDE_{ERROR|ASSERT}`, as printing to the screen and aborting the program is not always a good thing to do anyways (and is sometimes not even possible, in the case of a freestanding environment)