patch ptest-runner package in AGL to use Tim Orling's patch

We take the current way first and then see where things develop into upstream.
The ptest-runner variant looks to me better than the wrapper approach atm.

replying to myself... I'll assume that @Jan-Simon Moeller is ok with Tim's approach since he merged the patches into qa-testdefinitions, so will close this issue.

If we want to rethink this and switch to the Linaro way later, we'll track that as a separate issue.

Here's an AGL job for rpi3 using the Linaro ptest wrapper: http://lava.baylibre.com:10080/scheduler/job/59419

There's been some off-list discussion that Tim's approach of building LAVA support into ptest-runner itself is not the right way to go.  Rather, we should consider using the Linaro way of running all the ptests first, and then post-processing the output and sending to LAVA: https://git.linaro.org/qa/test-definitions.git/tree/automated/linux/ptest/ptest.yaml

Note that this has the benefit of being more like what users would do at their desktop, but has the downside of being all or nothing, since if the the ptest wrapper fails, the results will never be parsed by LAVA.

However, there does seem to be work in progress to have the ptest.py tool capture/send LAVA signals during the run, so we may be able to upgrade that when available.

@Jan-Simon Moeller what do you think?

Posted a series of patches to qa-testdefinitions that switch the ptest-wrapper from the custom wrapper to using the new LAVA mode: https://gerrit.automotivelinux.org/gerrit/#/c/18313/

Tested on an rpi3 in BayLibre lab.

Tested a local build of master branch on rpi3 in BayLibre lab.

minmal set: http://lava.baylibre.com:10080/scheduler/job/56313

full set: http://lava.baylibre.com:10080/scheduler/job/56314