Fix SIFT wrong octave indices + typo

[faberno]

Fixes #6173

• typo: changed sigmas to scales
• bug: wrong octave_indices if one octave wouldn't have keypoints, but the next one would

Description

Fixes the bug discovered by @longjoke. Extremas are saved in the extrema_pos list as seperate arrays for each octave. By using np.concatenate([np.full(len(p), i) for i, p in enumerate(extrema_pos)]), an array with the corresponding octave index of each extrema is created. If no extremas are found in an octave, nothing is appended to extrema_pos. This can to lead to wrong octave indices in a case where an octave has no keypoints, but the following does.
It's fixed by appending an empty array to extrema_pos, if no extremas are found in an octave.

I'm not sure, if it's possible that this case occurs.

For reviewers

• Check that the PR title is short, concise, and will make sense 1 year
  later.
• Check that new functions are imported in corresponding __init__.py.
• Check that new features, API changes, and deprecations are mentioned in
  doc/release/release_dev.rst.

[alexdesiqueira]

@faberno thank you for taking care of #6173! LGTM.

[faberno]

@alexdesiqueira I've also just added a small check to _compute_orientation if there are any keypoints in an octave, similarly as we have it in _compute_descriptor. This was not a bug, but the check can save some unnecessary computations

[rfezzani]

Excellent, thank you @faberno 😉

[grlee77]

@meeseeksdev backport to v0.19.x