'stronger' is a comparative adjective

Otherwise we won't get enough information from them.

There is only 1 step in this subprocedure now.

… and point to the most recent Debian bug that tracks the problem.

This partially reverts commit 1c31afe7.

Merge remote-tracking branch 'origin/bugfix/16942-spectre-v1-swapgs+force-all-tests' into devel (Closes: #16942, #16764)

Merge remote-tracking branch 'origin/feature/16981-tweak-gnome-app-menu' into devel (Closes: #16981)

This moves Calculator to Accessories and Terminal to System Tools,
by deleting the respective Include and Exclude lines in
/etc/xdg/menus/gnome-applications.menu.

This reverts commit 0c000cf8.

This was a workaround for https://bugs.debian.org/934483, which is now
fixed.

There's little harm in keeping this cleanup step, even though there's
a good chance it is useless… most of the time.

gcc-6 is not in Buster so there's no chance we've installed it
during the build ⇒ no need to deinstall it.

Ensure we don't install unwanted packages even if they become "Priority: standard" again (refs: #16949)

This is a follow-up to commits 368b038b and
46856445, that could lead to these packages
being installed again, e.g. in case of override discrepancy between the Debian
security archive and the regular Debian archive. This has bitten us a few times
in the past, e.g. with mutt. Let's avoid this biting us again for rpcbind, tcpd,
and exim4*.

The previous settings are a bit tough for machines that have 16GB of RAM or
a bit less: for example, one of the machines in my local Jenkins setup, where
I'm prototyping our next-generation CI hardware, has 32GB of RAM, and runs
2 Jenkins worker VMs, each having ~15GB of RAM. Since a few weeks I regularly
see builds failing because there's not enough free memory to start the Vagrant
build VM.

Back in November 2018, the chroot for our Buster-based image was way bigger than
when building Tails 3.x; with the "ram" build option, it therefore used more
memory in the build tmpfs; and with the "noram" build option, quite possibly
mksquashfs used more RAM as well. This probably explains why we had
to bump VM_MEMORY_BASE to 2048MB.

Since then, we've trimmed down our Buster-based image a lot, up to the point
where it's now 55 MiB smaller than 3.15. So there's a chance we can revert to
lower RAM requirements for the build VM. Let's try.

This partly reverts commit 09233368.

I've seen this step fail with:

      Could not find the 'The proxy server is refusing connections' error in the Tor Browser (RuntimeError)

… while the failure screenshot shows the expected heading.

It seems that when we list all headings initially, the one we're looking for may
not be here yet, so let's retry.

Merge branch 'feature/16912-move-greeter-to-main-git-repo+force-all-tests' into devel (Fix-committed: #16912, #12551, #16757)