Add support for aarch64

[andriish]

• Add support for aarch64
• Add CI that runs on Linux aarch64

Note: on aarch64, the float128 is equivalent to native long double and quadmath library does not exist.

[andriish]

Hi @scarrazza, could you, please, have a look?

Nowadays people make a lot of claims how beneficial it is to do calculations on ARM, but totally forget to invest in the support of software on ARM.

Best regards,

Andrii

[scarrazza]

Dear @andriish, thanks for this, it looks good to me.

[matthewfeickert]

This is great @andriish. With this we can also get Linux aarch64 (and I think also macOS arm64) builds working for the conda-forge distribution: https://github.com/conda-forge/qcdloop-feedstock

(c.f. conda-forge/qcdloop-feedstock#8).

edit: @andriish Out of curiosity, have you patched aarch64 support into other libraries that use libquadmath? I know that this is also an issue with ninja (https://github.com/peraro/ninja) (c.f. hep-packaging-coordination/packaging-hep-simulation-stack#3)

[matthewfeickert]

@andriish Doesn't this assume only the existence of x86 and aarch64? Won't this break ppc64le builds now?