Use nearest-neighbors interpolation in plot_component

[tsalo]

Closes #1095.

Changes proposed in this pull request:

• Switch from the default 'continuous' resampling approach to 'nearest' in plot_component's plot_stat_map call. This should stop non-diagonal images from being resampled to have low, non-zero values outside of the brain.

[tsalo]

From Logan's 8-echo dataset.

Before:

After:

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 89.91%. Comparing base (12aea7f) to head (d8df3e6).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1098   +/-   ##
=======================================
  Coverage   89.91%   89.91%           
=======================================
  Files          26       26           
  Lines        3621     3621           
  Branches      629      629           
=======================================
  Hits         3256     3256           
  Misses        214      214           
  Partials      151      151           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[handwerkerd]

For the three times vmin is set to the 2nd percentile, you would not need to set resampling_interpolation="nearest" and "continuous" will result in nicer looking images.
For the one situation where vmin isn't set, setting vmin=np.min(stat_img.get_fdata()) might give slightly nicer visualizations, but setting interpolation to "nearest" is also fine.

[tsalo]

Good point. I dropped the change for plot_stat_map calls outside of plot_component.

[handwerkerd]

LGTM

I also realized my suggestion for adding vmin to stat_map figures is a bit more complex, since there are positive & negative values. It is solvable, but this solution is fine.

[tsalo]

Yeah we'd need to use threshold instead of vmin, but I played around with that and had some issues, so I think this is fine for now. If we end up overlaying our stat maps on other images (e.g., mean OC, T1w) then we might need to revisit.

[tsalo]

I'd like to make a release with this change before we merging #1064. @handwerkerd any objections?