Set up a functioning integration test

[maehler]

Things that need addressing:

• Fix config files Fixed in 941bcab
• Picking a set of input reads that is not too big
• Making sure that the resulting alignments make sense. This is especially the case for pindel since it currently fails due to the insert size from picard being shorter than the read length.
• Set up minimal needed references files. For example, cnvkit crashes at the moment due to an empty panel of normals (at least I think this is why).
• Drop the Github Actions integration test. Removed in 2375979
• Using webhooks, set up an integration test somewhere else using a real dataset.
• Follow Twist Solid and drop conda support. In a way, this was addressed in 2375979, but I will leave the conda environment yaml for copying the result files in case anyone would want to try their luck using conda. Removed all mentions of conda in b30e8c0.

The branch associated with this work is fix-integration-test-action.

[maehler]

Based on experiences with the Twist Solid pipeline, this might be very hard (impossible?) to fit into something that can be run as a part of Github Actions. It is mostly the CNV part that is causing problems where a subset of data simply isn't sufficient to capture the intricacies of a full dataset. A better solution would probably be to set up testing on a full dataset somewhere, making use of Github webhooks.

[maehler]

Now, the "only" thing left is to set up an integration test with real data. This will not be a quick fix, so I will revisit this at a later stage.