Failing on WASM and 32-bit builds

[JensMunkHansen]

On WASM, the non-intrinsic version pass with -02, but fails with -O3. I tried running it on x86, i.e. -m32. Here the non-intrinsic worked, but some patching is needed to get the bmi version to work.

Do you have any plans on making it work on WASM. For WASM morton codes is super relevant, since the operate with a 64 kB cache, which requires small search structures.