Split `-f` flag into separate options: include results from failures vs exitcode handling

[smarr]

Currently, the -f both means:

1. that we include data from failing runs, which can be caused either by benchmarks aborting or because the validation of results fails or output indicates other issues

2. that the rebench exit code indicates no failure

We may want to separate these concerns out.

It might be useful to have a --dont-fail option, just to keep the exit code 0.
Where --faulty is only about including data from runs that indicate some issue.

Requested by @OctaveLarose