Implement FEATURE suppress-bindist-buildpkg

[Kangie]

This feature combines with buildpkg to enable the installation of packages with "RESTRICT=bindist" without the creation of a binpkg.

This should help binhost builders improve the library of pre-packaged software without running the risk of accidentally (bin)packaging a dependency that we can't legally distribute (or don't want to pay royalties on!)

Current Behaviour:
FEATURES=buildpkg + RESTRICT=bindist
=> Package is installed, binpkg created
FEATURES=buildpkg + RESTRICT=bindist + ACCEPT_RESTRICT=-bindist
=> Package is masked, cannot be selected

New Behaviour:
FEATURES=buildpkg + RESTRICT=bindist
=> Package is installed, binpkg created
FEATURES=buildpkg suppress-bindist-buildpkg + RESTRICT=bindist
=> Package is installed, binpkg creation stage skipped
FEATURES=* + RESTRICT=bindist + ACCEPT_RESTRICT=-bindist
=> Any combination of features and the above are masked

Bug: https://bugs.gentoo.org/542480

[eli-schwartz]

If your only reason to do this is that the Gentoo repository doesn't support it anymore then just don't do it?

I'm generally quite opposed to bumping versions just for the heck of it, without actually benefitting from doing so. No, dropping it from a CI matrix doesn't count.

Consider also, that dropping python versions means non-Gentoo entities who also use portage, may have their workflow broken. e.g. stable LTS system builders, who should at least be given the opportunity to use any non-EOL version. Also uhhhh the elephant in the room being Google, who have interesting perspectives on which version of Python to support. :)

If you can't even muster up a single code change to make as a result of dropping 3.9 support then it indicates your whole claim of dropping support is entirely hollow.

...

Also, please propose completely unrelated distinctive changes in different PRs, for better reviewability.

[Kangie]

Well, since you got here before I had a chance to comment:

I've bumped the minimum requirement to py3.10 as this change fails with py3.9 (I used match in a test), which wasn't picked up until CI runs because I don't develop against EoL'd (or almost EoL'd) Python.

It wouldn't be hard to rework the code to work with older python, but I also don't see much point in making such a change if we're not actually sure that we want py3.9 anyway. Since we're reviewing changes, I opted to drop the older version (which was privately discussed earlier as a possibility). I'm equally happy to just use if/else if we really want old python support.