Tool testing improvements

[mvdbeek]

These are a couple of things I found while implementing tool tests:

• [x] tests only run of the test data is available from where the tests are triggered (this should be fixed with a galaxy-lib update containing https://github.com/galaxyproject/galaxy/pull/7160)
• [x] tests seem to require an admin API key, that shouldn't be necessary with https://github.com/galaxyproject/galaxy/pull/7160
• [x] We create a new history for every test, even if this test is for the same tool
• [x] We just raise JobOutputExceptions, these need to be displayed properly
• [x] We should have some reporting, ideally re-using some things from planemo (manual planemo test_reports tool_test_output.json works fine)
• [ ] Optionally delete test histories after run
• [ ] Skip tests that use reference data
• [ ] Move to next test when job becomes paused
• [ ] Expose test timeout as command line option

[selten]

@mvdbeek would it also be an idea to add the ability to remove the histories after all tests are finished (or at least as a command line option)?

[mvdbeek]

Yep, that would be a good option. I've added this along with some other things we've recently come across

[selten]

shed-tools test should have an option to exit with exit code 1 in case of failed tests. For me it seems to exit with exit code 0 even if it has failed tests.