Wrong number of executed tests reported in output

[teojgo]

The total number of (executed)/(total) tests is wrong in the output, e.g:

[       OK ] (720/717) SarusMpichOSUTest on dom:gpu using builtin [compile: 0.008s run: 74.909s total: 74.941s]
[----------] all spawned checks have finished

[teojgo]

Somehow that failure of the CudaFortranCheck is printed twice and the test is counted twice as well.

[teojgo]

The problem happens only with the async execution policy. There is a bug on:

https://github.com/eth-cscs/reframe/blob/5f3c9192ba2494bd310b6b3c3c1bba61d18c0a4b/reframe/frontend/executors/policies.py#L412

which makes the check to be rescheduler when it fails. The line should be:

if not (task.failed or task.skipped)