support dense head surfaces in `plot_alignment`

[drammock]

There does not appear to be a way to specify whether to use the sparse or dense head in mne.viz.plot_alignment. According to @larsoner this used to be possible with the now-deprecated plot_trans function. Requesting an enhancement to allow this via plot_alignment.

The loading of the head surface is handled here so presumably that's where code to allow user-specified surface density would get plugged in.

[agramfort]

+1

PR welcome

[larsoner]

We could make things like surfaces='sample-head-dense' would work by looking for .surf and .fif extensions, and loading the surface or BEM surface, respectively. WDYT?

[agramfort]

I would use head-dense.

cf doc:

surfaces : str | list

Surfaces to plot. Supported values: ‘head’, ‘outer_skin’, ‘outer_skull’, ‘inner_skull’, ‘brain’, ‘pial’, ‘white’, ‘inflated’. Defaults to (‘head’,).

[larsoner]

That's doable but a departure from what is done now. The existing logic actually assumes 1) you've given something like the entire filename as it is in $SUBJECTS_DIR/$SUBJECT/bem/, and 2) you're loading a .surf file (or one of the surfaces in your already-loaded bem). So in cases where people had sample-head-dense.fif, it wouldn't load even if supplied head-dense.fif because it doesn't prepend the sample- at the beginning. We could add this prepending behavior (as well as the ability to also look for and load .fif) -- is that what you're thinking?

It looks like the existing code also doesn't check to see that all surfaces supplied were found (so surfaces='foo') won't raise an error) so we should fix that while we're at it.

[agramfort]

sample-head-dense.fif is a standard name. I think we can guess it.

+1 for the fix