feat(Build): Build manylinux_2_28 wheels

[reneleonhardt]

Moved 31* out of manylinux2010 and build_wheels into a new manylinux_2_28 job.
Moved PYTHON_VER to .python-version.

Not built anymore:

• cp310-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686
• cp310-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64

Included in manylinux_2_28 now:

• cp310-manylinux_2_17_x86_64.manylinux2014_x86_64 (old)
• cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.manylinux_2_28_x86_64 (new)

[reneleonhardt]

Could we build against aarch64 too? Then I could skip those in build_wheels too, names are the same:

• cp310-manylinux_2_17_aarch64.manylinux2014_aarch64
• cp310-musllinux_1_1_aarch64

[mrabarnett]

How much demand is there for that?

If you can provide the relevant changes and confirm that it works, then I don't see why not.

[reneleonhardt]

You developed the best regex engine for Python as a C extension, so aarch64 builds will be useful to the growing ARM cloud offerings.
Please try out yourself.

Did you ever think about a Java port or JNI / Foreign Function use?

[mrabarnett]

Hello again!

If you're able, could you have a look at mail.yml again. It needs updating, and I've given up trying to understand how to fix the Linux build problems.

[reneleonhardt]

I'm not sure I'm following, you didn't merge my work for 2 years, what's there to look in main.yml?
It wouldn't matter anyway if contributions are never merged.