Do not fail series if the test fails on the base

[bonzini]

When a test is failing on master, currently it will end up failing on all submitted patches too. We should instead mark the results as BLOCKED rather than PENDING in the series; the test can be triggered manually on the series that fixes the breakage.

This way: 1) we can make the "T" icon yellow instead of red; 2) we don't inundate the mailing list with failures; 3) we can easily search the affected patches and retrigger testing on them (with "blocked:testing" as a search term).

[famz]

Related: #32

[famz]

If all series are still tested regardless of their base (master) failure, the "fix" for the base breakage can be verified without human involvement.

All that is required is to turn "errors" into "warnings" when failures are also seen on a series "base", I think.

[bonzini]

True. So another possibility is to add a base_status to the testing result data, and use it to distinguish error (red) from warning (yellow) in the dashboard.