FIX/DOC/ENH: inaccurate reconstruction of times from matlab files

[dengemann]

Somethings is slightly off when reconstructing epochs.times from matfiles. The bounds and offsets are not correct. This gets extreme as strongly decimated files are read. One central issue here is that the preprocessed epochs were already decimated, potentially with a different method, e.g. linspace(min, max, n_samples). My suspicion is that either the sampling frequency reported in the matfiles is not correct or the times vectors. Some discussion on this took already place here:

#9 (comment)

#9 (comment)

#9 (comment)

The task is to figure out what needs to be fixed, the data descriptors -- on reading or MNE-Python .times reconstruction.

cc @Eric89GXL @jasmainak etc.

[larsoner]

Does 1. / (times[1] - times[0]) == sfreq?