Always raise, don't swallow exception, if exception is KeyboardInterrupt in ProtocolUnit.execute

[dotsdl]

This PR is an attempt to address openforcefield/alchemiscale#132.

@mikemhenry raised a good point that for the case of a KeyboardInterrupt, even if raise_error = False for ProtocolUnit.execute, we probably always want to raise.

Is there any downside to this?

[codecov]

Codecov Report

Patch coverage: 100.00% and no project coverage change.

Comparison is base (2028a63) 99.12% compared to head (9146ea8) 99.12%.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #215   +/-   ##
=======================================
  Coverage   99.12%   99.12%           
=======================================
  Files          36       36           
  Lines        1829     1831    +2     
=======================================
+ Hits         1813     1815    +2     
  Misses         16       16           

Files Changed	Coverage Δ
gufe/protocols/protocolunit.py	97.24% <100.00%> (+0.03%)	⬆️

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[dwhswenson]

My only concern would be if INT is being used elsewhere, e.g., by a queueing system, in which case we'd want to catch it and terminate cleanly. (Which also might be an option here, anyway.)

The queueing systems I know send TERM or USR2, followed by a KILL; I'm not aware of one using INT, but I don't claim to know them all!

If we ever handle mid-unit restarts, we'll need to handle KeyboardInterrupt with a little more logic, but that's easily added. Otherwise, seems reasonable -- either needs a test or exclude from coverage.

[dotsdl]

@dwhswenson thanks for this! I've added an explicit test for the new behavior.

As for the concern you raised about queueing systems, I think you're right that SIGINT generally isn't used for terminating jobs, so I think we're still good here.