Refactor nwb extractor frames and tests

[h-mayorquin]

This is the first attempt at getting correct frames in nwb_extractor. This branch should be tested in nwb-conversion-tools to see if everything works as we expected over there. I refactored the function and added a test to see that it works as expected.

I intend to add a test to support loading TwoPhotonTimeSeries with channels but that's not the priority right now.

[h-mayorquin]

This is the PR in nwb-conversion-tools:

catalystneuro/nwb-conversion-tools#547

[codecov]

Codecov Report

Merging #136 (7903b2c) into master (a6c08b8) will increase coverage by 1.67%.
The diff coverage is 86.04%.

@@            Coverage Diff             @@
##           master     #136      +/-   ##
==========================================
+ Coverage   66.38%   68.06%   +1.67%     
==========================================
  Files          31       31              
  Lines        1919     1926       +7     
==========================================
+ Hits         1274     1311      +37     
+ Misses        645      615      -30     

Flag	Coverage Δ
unittests	68.06% <86.04%> (+1.67%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
...tractors/extractors/nwbextractors/nwbextractors.py	47.73% <81.81%> (+16.10%)	⬆️
src/roiextractors/testing.py	100.00% <100.00%> (ø)

[h-mayorquin]

@CodyCBakerPhD
I corrected to take into account the transposition and the fact that the TwoPhotonSeries does not really have channels as per the discussion in:
NeurodataWithoutBorders/nwb-schema#522

Also, I realized that most of the other get_frames functions do not behave like numpy when indexed with a singleton. I will open another issue for that.