[ENH] Creates plot_fsvertex() function

[rmarkello]

For plotting vertex-level data using pysurfer. Also refactors plot_fsaverage() so that after initial input formatting it calls plot_fsvertex() under the hood.

@VinceBaz since you've worked on plot_fsaverage() before maybe you could take a look at this (to make sure it makes sense) and ensure I didn't screw up anything too dramatically with my refactoring? If you're not working on anything that would be able to test the new plot_fsaverage() at the moment then no worries.

[codecov-io]

Codecov Report

Merging #95 (34834c1) into master (afa3af3) will increase coverage by 3.03%. The diff coverage is 32.78%.

@@            Coverage Diff             @@
##           master      #95      +/-   ##
==========================================
+ Coverage   64.65%   67.69%   +3.03%     
==========================================
  Files          23       24       +1     
  Lines        1610     1681      +71     
==========================================
+ Hits         1041     1138      +97     
+ Misses        569      543      -26     

Impacted Files	Coverage Δ
netneurotools/plotting.py	32.82% <11.76%> (+26.24%)	:arrow_up:
netneurotools/tests/test_plotting.py	81.08% <81.08%> (ø)
netneurotools/modularity.py	59.55% <0.00%> (-4.50%)	:arrow_down:

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update afa3af3...a903253. Read the comment docs.

[VinceBaz]

It's a nice enhancement @rmarkello and sorry for the late reply! Overall I don't see anything that might be problematic and you can go ahead and merge the branches.

A very minor issue might however be how you handle vmin and vmax in plot_fsaverage() and plot_fsvertex(). Basically, I don't think that handling vmin and vmax in plot_fsaverage() is necessary since you handle these two parameters in plot_fsvertex() (Unless there's a clear reason to do so?). Also, I would argue that using np.nanpercentile() instead of np.percentile(np.nan_to_num()) , assuming you use this function to avoid NaNs, would be better: np.nanpercentile() would completely ignore nans, which might be what the user wants if they rely on a mask to hide their NaN values in the array vector.

[rmarkello]

@VinceBaz thanks for this! Sorry for my delay in getting back.

(1) I handle vmin and vmax separately because if you're using plot_fsaverage() I want the percentiles to be based on the parcellated data, not the dense data (i.e., I want to avoid influencing the colormap range because some parcels might have more vertices than others). It's not elegant, but I couldn't think of another way to do it... Let me know if that makes sense to you, though. (2) np.nanpercentile() is definitely what I wanted, thank you ! 🙏 I will make the relevant change (and fix the failing tests) soon.

[rmarkello]

Getting those tests to pass was painful but it's done! 🙌 We are now officially testing (some) of our plotting functions, and adding tests for the others should be much easier to do in the future :grin:

Gonna go ahead and merge this in, finally