Add volume mapper choice notes

[banesullivan]

The 'smart' mapper chooses one of the other available mappers based on rendering parameters and available hardware. I added some notes on how that decision is made.

I also added a waring about the fixed_point mapper and how it CPU-based

Addition notes not included in docs

I noticed on my machine today that when datasets have NaN values, the 'fixed_point' mapper might work best, but the 'smart' mapper was choosing something different. I added these notes to the docstring to help others. I do not want to change the default behavior as the 'smart' mapper usually chooses the best mapper for the given hardware and autoselecting 'fixed_point' in the presence of NaN values could fail for some hardware.

I'm seeing this issue on my M2 Mac but not my Linux machine with an NVIDIA graphics card

import pyvista as pv
from pyvista import examples
import numpy as np

volume = examples.download_knee_full()

volume['SLCImage'] = volume['SLCImage'].astype(float)
volume['SLCImage'][volume['SLCImage'] < 40] = np.nan

p = pv.Plotter()
p.add_volume(volume, cmap="coolwarm", mapper='fixed_point')
p.show()

[codecov]

Codecov Report

Merging #3824 (30440d7) into main (880d172) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main    #3824   +/-   ##
=======================================
  Coverage   94.18%   94.18%           
=======================================
  Files          86       86           
  Lines       18914    18914           
=======================================
  Hits        17815    17815           
  Misses       1099     1099           

[banesullivan]

I actually changed this up a bit and removed the original NaN note. That NaN issue seems to be specific to arm-based Macs and the OpenGL implementation on that hardware. I don't really want to point people to using the fixed_point mapper since its performance is quite poor (being CPU-based)

[tkoyama010]

LGTM