We decided that developers who want to do different things in the hook
when running in test suite can check for the "unmount" and
"autotest_never_use_this_option".

refs #19237

The original child method already has regex support, we don't need to
implement it ourselves.

The step was failing on my system because a few pixels are missing on
the bottom of the screen keyboard.

Those are printed repeatedly so often that it's too much noise.

To be able to find the method definition via IDE navigation.

This also drops some unused methods.

The test suite changes of this commit require the updates to
tails-autotest-remote-shell from commit
"Test Suite: Fix Dogtail not working when running as Debian-gdm user"
to work. That causes some scenarios of the
"Upgrading an old Tails USB installation" feature to fail, because the
old version of Tails used by that test does not have the updates to
tails-autotest-remote-shell yet.

To avoid that, lets wait until the next release and then revert this
commit again.

In one scenario, the next step after this one was to shut down, which
caused a race: Because the step did wait until deactivation completed,
sometimes the feature was not deactivated yet, causing it to be active
again after reboot.

Closes #19420

Clicking on screenshots is flaky, both locally and in Jenkins. Using
Dogtail should be more robust.

Changing the config of 'dogtail.config' doesn't have any effect on the
config of 'dogtail.tree'. To do that, 'dogtail.tree.config' needs to be
changed.

We decided to add a --extra-boot-options to run_test_suite instead.

This allows us to inspect the journal and do other things as root when
debugging test failures in the VM.

When importing "dogtail.tree" when the AT SPI bus was inaccessible, the
Python process was terminated via SIGTRAP:

    (.:3116166): dbind-ERROR **: 14:35:52.927: AT-SPI: Couldn't connect to accessibility bus. Is at-spi-bus-launcher running?

That was hard to debug because we didn't print the stderr of the failed
Python command. With this commit, failures like the one above should be
a lot easier to debug.

Handle exceptions during JSON decoding by returning the exception to the
caller instead of letting the Remote Shell exit abruptly which makes
things hard to debug.

In other steps (e.g. "the file exists") filepaths are quoted, so lets do
that consistently

Sometimes importing "dogtail.tree" failed with:

    dbind-ERROR **: 11:16:00.750: AT-SPI: Couldn't connect to accessibility bus. Is at-spi-bus-launcher running?

Setting the AT_SPI_BUS_ADDRESS environment variable fixes that.

Instead of converting all input/output from/to string.

... for easier debugging. The clipboard is only shared when connecting
to the SPICE display with e.g. virt-manager to debug things, not when
the VNC viewer is connected via --view or --view-interact.

The blue ansi color code (34) is often hard to read in terminals with a
dark background. Lets use cyan instead, which should be readable with
both bright and dark background.

Since Ruby 2.1, when an unhandled exceptions is printed, any previous
exceptions at the time the exception was raised are also printed. That's
useful to for debugging to figure out the exact cause of an exception.

Cucumber does not print previous exceptions, so we extend it's
print_exception method to also print any previous exceptions.

This allows to do different things when early-patching test suite runs.
Developers who don't need this feature can just create a symlink from
hook-test-suite to hook.

We still also set early_patch=umount to also support early-patching ISOs
which don't have this commit yet.

refs #15211

get_field('sensitive') returns either 'True' or 'False' which both
evaluate to true.

Also, to ensure that pressing Return in the correct entry after a
grabFocus call, we should check for 'focused' not 'sensitive'.

Even if it's possible that chutney failed because of the existing data
directory, it's still helpful to print the error that caused chutney to
fail.