LAVA callback implementation clean-up

[gctucker]

The LAVA callback code in kernelci-backend has been evolving gradually to cope with corner cases and various issues with the data sent from LAVA. We should now take a look at it again and make a cleaner implementation based on what we've learned. This should also help fixing the remaining known issues such as test cases having the whole boot log associated with them.

• JSON callback file that shows the issue described on baseline.login result doesn't match job log #244
• JSON callback file that shows the issue where the first test case has the full boot log
• unit test that takes JSON fragments to reproduce these issues and gets a data structure from the code with the sequence of test cases and their associated log lines
• other unit tests can be added to cover the part that creates the tree of test cases in sub-groups
• fix implementation to make the unit tests pass, with potentially new functions to replace the current implementation
• verify that the new code doesn't show the issues any more using the full JSON callback data files
• test on staging.kernelci.org for a full integration test in the KernelCI pipeline

[mgalka]

lava-json-.zip
JSON callback files causing issues:

• lava-json-jetson-tk1.json - Whole bootlog attached to a test case
• lava-json-meson-gxbb-p200.json - no log lines attached to the emerg test case

[gctucker]

@mgalka Thanks, it would also be useful to have a JSON callback dump to reproduce the issue described in #244, essentially with a kernel "bug" that triggers the LAVA state machine to get to auto-login before reaching an actual login prompt. If that issue is already fixed in the current backend, please put a comment on it and close it. Having the JSON in any case will help for regression testing, especially as the code is going to be refactored.

[mgalka]

GH_244.json.txt

JSON callback file that reproduces #244
File extension changed to .txt due to GitHub attachment name policy.