FIX set default imshow interpolation to nearest for composite.add_data

[mvdoc]

The default image interpolation in matplotlib is now antialiased. This can cause issues when plotting data containing NaNs with cortex.quickflat.make_figure. For example, when setting different dpi levels, the images can look wildly different. To fix this issue, this PR sets the default interpolation of the imshow in composite.add_data to nearest. (Thanks to @alexhuth for help debugging this issue!)

The following code demonstrates how antialiased interpolation can cause issues:

import cortex
import numpy as np
import matplotlib.pyplot as plt

vol = cortex.Volume.empty('S1', 'fullhead', vmin=0, vmax=2) + 1
vol.data[np.random.rand(*vol.data.shape) > 0.6] = np.nan

img = cortex.quickflat.make_flatmap_image(vol, thick=1, nanmean=False, height=800)

fig, ax = plt.subplots(1, 1)
ax.imshow(img[0])
fig.savefig("test-dpi-100-make_flatmap_image.png", dpi=100)
fig.savefig("test-dpi-300-make_flatmap_image.png", dpi=300)

Setting interpolation='nearest' fixes the problem:

import cortex
import numpy as np
import matplotlib.pyplot as plt

vol = cortex.Volume.empty('S1', 'fullhead', vmin=0, vmax=2) + 1
vol.data[np.random.rand(*vol.data.shape) > 0.6] = np.nan

img = cortex.quickflat.make_flatmap_image(vol, thick=1, nanmean=False, height=800)

fig, ax = plt.subplots(1, 1)
ax.imshow(img[0], interpolation="nearest")
fig.savefig("test-dpi-100-make_flatmap_image.png", dpi=100)
fig.savefig("test-dpi-300-make_flatmap_image.png", dpi=300)