ENH: Measure head motion from cHPI

It might be helpful to know how much head motion there was in a scan when cHPI was used, either for QA/reporting or to help decide whether to use movement compensation in maxfilter/maxwell_filter.

If we do something like what happens for fMRI, the steps would be (roughly) to 1) use chpi.read_head_pos or chpi._calculate_head positions to get the quaternions, 2) convert the translations to mm (as in viz.plot_head_positions), 3) convert the rotations to radians to degrees*, 4) concatenate translations and rotations into an array, 5) do a diff, 6) take the euclidean norm at each measurement (e.g., second), and then 7) take the mean and standard deviation of the norms to get motion per second.

*The logic of treating degrees of rotation as mm is borrowed from afni, see e.g. https://afni.nimh.nih.gov/afni/community/board/read.php?1,82641,82645#msg-82645 .

And I made a gist: https://gist.github.com/jhouck/b0e93a60f17c708c5a342dce354551e2

It might make sense to use some standard time with chpi._calculate_head positions, like setting both t_step_min and t_step_max to 1.0, but maybe that would be left to the user.

Thoughts?

Start by reviewing chpi.read_head_pos and chpi._calculate_head, then compare the translation handling in viz.plot_head_positions with the proposed gist. Check how maxfilter/maxwell_filter currently consume head-position data. Done should include an agreed API and validated motion-per-second statistics, but the issue does not name tests or implementation files.