build-style/python3-pep517: use a generic glob for wheels

[tornaria]

We replace the current glob of "dist/${wheelbase//-/_}-${version}-*-*-*.whl" for a much simpler dist/*.whl. The former is inconvenient since wheelbase="${pkgname#python3-}" is most of the time correct but often not, and fixing that in a different way seems more complicated than this solution.

Since we only run python -m build once, we should have only one wheel, so trying to be more specific doesn't seem useful. Nevertheless, this can still be overriden via make_install_target as before but hopefully it won't ever be necessary.

Note that several packages that currently need to set make_install_target for this purposes will now work out of the box.

If this is accepted I can have a look at those pkgs and clean up the override once it's no longer necessary.

Testing the changes

• I tested the changes in this PR: YES

I built all 150 packages that use pep517 as obtained by:

$ git grep -l style=.*pep517 srcpkgs/ | cut -d/ -f2 | wc -l
150

All of them succeed except for two that fail for unrelated reasons (python3-xcffib and synapse).

I also included a minor unrelated change: do not compile bytecode in do_install() since it will be removed in post_install().

CC: @ahesford @icp1994

[ahesford]

I have made a couple of changes:

1. A squash commit refactors do_check: a) to short-circuit checks when pytest is not installed, avoiding the need to nest the bulk of the function inside a conditional; b) to avoid setting make_install_target in do_check when it is undefined in the template, deferring the default assignment to do_install.
2. Remove the now-superfluous make_install_target in all PEP517 templates.
3. Fix the build of synapse and (by updating) pythone3-xcffib.

Please confirm that you are OK with these changes, and we can squash down my alterations before merge.

[icp1994]

For synapse, there is a update PR #43815 which includes the raised setuptools_rust bound from upstream

[ahesford]

Thanks; I dropped the synapse fix.

[tornaria]

LGTM, thanks! I made a minor suggestion so the style in do_install() matches the style in do_check(). I prefer the other way (setting default values at the top of the function) since that acts as documentation and makes it much more clear what are default values, etc. and it's a style used throughout.