Add masking for feature detection

[3nol]

I created this PR as a follow-up for PR#55 since no further progress has been made. I based my implementation on @piercus's attempt, so he receives partial credit.

I only implemented the first part of the above PR (feature masking, not background stitching etc.). But I think this is already very helpful for stitching.

This PR adds:

• The --masks cli parameter (as mentioned, only feature masks for now)
• Change to the ImageHandler to allow storing mask file names, as well as corresponding accessors
• An additional test for feature masking in test_detector.py

[piercus]

Looks great ! Thanks for the follow-up !

I feel focusing on features only is the best choice !

[lukasalexanderweber]

Thanks for keeping up on this! I'm leaving into holiday and need some time to review this, I hope its fine that I check back to you when I'm back.

[lukasalexanderweber]

During the review of the PR I noticed that the ImageHander has too much state and also a bug (817c3d0). I would like to refactor/fix this first, because I think that then not so much code has to be inserted into the image handler, which is already too complicated for my taste. I hope this is fine for you.

[lukasalexanderweber]

what would be helpful in the meantime is a test where the full stitching worklow is run with and without feature masks and we see that different results are achieved. You could place the images here, we could then present the feature in the tutorial and download it from there for out tests.

[3nol]

Thanks for looking into this again! Yes, that sounds like a good idea. After all, using masks is a very user-facing feature, and a tutorial with working examples might be very helpful.

I will add a test with the full stitching workflow (and example images) in the next couple days.

Edit: I am sorry, I won't be able to finish this soon. If anyone wants to help, please feel free.