stem_activations needs documentation

[bmcfee]

Two points:

• mtrack.stem_activations is a list of lists, but the docs say it's an ndarray
• The shape does not line up to the number of stems; it's off by one. It looks like the first column is reserved for a time index, not activations.

The first point is an easy fix.

The second point is confusing, and it's generally not good style to mix indexing/addressing (ie timestamps) with observation data. (If you do so, it should definitely be documented.) I recommend refactoring this so that the time index is stored separately.