Remove unused np.where from rmsimage.py

[AlexKurek]

Since np.where turns out to be so slow (#311), it would be good to remove it where it is not needed.
.gaul files are identical, but currently the test will fail due to #314.

[gmloose]

The function numpy.where() is also used in several other places, a quick grep reveals. I didn't look if some of these calls can also be removed safely, but it might be worth a check. If so, I have no objection if you extend this MR to cover the removal of all unnecessary calls to numpy.where().

[AlexKurek]

At the other places np.where has an actuall effect and is needed. Except this one:
https://github.com/lofar-astron/PyBDSF/blob/master/bdsf/rmsimage.py#L961
but I think there may also be a bug in this block (#317), so I think its safer to leave it for another PR.

[AlexKurek]

Generally, np.where is redundant for boolean masking, checking element existence, or counting matching elements. Instead, directly use boolean indexing (arr[mask]), np.any(), or np.count_nonzero(). This avoids unnecessary memory overhead and improves speed.