BUG: Safe hasattr for PyQt5

[larsoner]

Over in enthought/mayavi#1180 I am working on VTK 9.2.2 fixes, and I am getting a segfault in CIs during GC:

https://github.com/enthought/mayavi/actions/runs/3222141217/jobs/5270891488

Locally I don't get a segfault but rather a slightly nicer error message:

Exceptions caught in Qt event loop:
________________________________________________________________________________
Traceback (most recent call last):
  File "/home/larsoner/python/pyface/pyface/ui/qt4/action/action_item.py", line 379, in _qt4_on_destroyed
    if hasattr(self.control, "_tool_instance"):
RuntimeError: wrapped C/C++ object of type QAction has been deleted

This PR fixes this by catching this RuntimeError during hasattr. Locally at least the Mayavi tests pass with PyQt5 with this change.

[corranwebster]

Thanks - this looks like a reasonable fix.

[larsoner]

There was a timeout on one run after 360 minutes (!) that's hopefully unrelated, so rather than pushing an empty commit (to get CIs to run again) I added a timeout-minutes: 20 to the CI run. Looks like all others were < 5 minutes so this seemed like a reasonable buffer against slow downloads and such.

[corranwebster]

Yes, it is almost certainly orthogonal - I haven't seen that sort of error in the CI runs, but I will keep an eye out for it in the future.

I think this should be good to merge.