Move ORB to SIFT feature detection

[khufkens]

SIFT is now off patent and should be better than ORB feature matching for the template matching routine

[derrickmuheki]

I've just had a quick try with this and i get the figure attached here.
template_matched.pdf
using the figures below:

It seems there are diagonal matches detected even though that shouldn't be the case. Perhaps, I need to check my code more. i used the Brute Force Matching with SIFT described here: https://docs.opencv.org/4.x/dc/dc3/tutorial_py_matcher.html

[khufkens]

There is a lot of filtering afterwards, also note that the template does not match the target in content. This is a different format. Note the switch in the label on the left! These things are sensitive to these changes. You should also remove or trim the template to be as small as possible (not include the black border - the same goes for the target page but I think this is covered in the routine).

[derrickmuheki]

**Update. I've over the past weeks made some good progress on the transcribing and will update the github in the next days after cleaning up. I used k-means clustering for ensuring the bounding boxes are placed in the right locations in the output excel file and it works well for the few sheets that I've tested it on. We'll see how it performs on more different images.