Consolidate some legacy test constructs

[dmnks]

• Merge RPMDB_INIT into RPMTEST_SETUP
• Rename RPMTEST_SETUP to RPMTEST_INIT
• Add RPMTEST_SETUP as a wrapper for AT_SETUP (for consistency)
• Drop run()
• (stretch goal) Define wrappers for the RPM binaries (rpm, rpmbuild, rpmsign, ...) which would be used instead of runroot*() (optimization: don't run a Bubblewrap container if $RPMTEST is /)
• ...and finally eliminate the need for explicit RPMTEST_INIT at all

Thanks @pmatilai for the ideas! 😄

[dmnks]

* Perhaps `RPMTEST_SETUP` isn't very descriptive, choose a better name? Its purpose is to initialize a writable tree and mount it at `$RPMTEST`

Hmm, actually this is probably a good name after all 😄 It literally says "setup RPMTEST (directory)"...

[pmatilai]

The problem with RPMTEST_SETUP name is that it seems to relate to AT_SETUP when it doesn't. Especially as we have RPMTEST_CLEANUP which relates to AT_CLEANUP, and one day there may be a reason to wrap AT_SETUP similarly.

Maybe RPMTEST_INIT?

[dmnks]

Yup, good point. It could be made into an AT_SETUP wrapper and just take some kind of argument that says "I need write access to /" but that would require a lot of (even if automated) search & replace across all the test files...

In any case, RPMTEST_INIT works for me.

[pmatilai]

After it's renamed, it may not be a bad idea to actually add RPMTEST_SETUP that does, for now, nothing more than wrap AT_SETUP. Just to make the whole thing more consistent.

[dmnks]

OK, I've updated the description accordingly, thanks!

[pmatilai]

Yay, we got to drop run() 😄

[pmatilai]

Seems this has turned into halfway rewrite by now 😂 but it'd seem strange not to include #3542 in this as well...

[pmatilai]

I think we'll consider this step done now. There are other legacy constructs to deal with but those are better dealt with on in their own tickets.