Multi-gate and multi-trigger t_starts in SpikeGLX are incorrectly set to 0

[h-mayorquin]

As discussed with @samuelgarcia the starting time of the segments in SpikeGLX are set to 0 here:

python-neo/neo/rawio/spikeglxrawio.py

Lines 231 to 238 in 0ae6e76

	# deal with nb_segment and t_start/t_stop per segment
	self._t_starts = {seg_index: 0.0 for seg_index in range(nb_segment)}
	self._t_stops = {seg_index: 0.0 for seg_index in range(nb_segment)}
	for seg_index in range(nb_segment):
	for stream_name in stream_names:
	info = self.signals_info_dict[seg_index, stream_name]
	t_stop = info["sample_length"] / info["sampling_rate"]
	self._t_stops[seg_index] = max(self._t_stops[seg_index], t_stop)

But in the metadata there is a firstFrame value:

https://billkarsh.github.io/SpikeGLX/Sgl_help/Metadata_30.html

firstSample=779283
All acquired samples are assigned an index number; the first acquired during the run is sample 0. This value is the index number of the first sample recorded in this file.

This is important so the correct times can be propagated to downstream applications.

We have data on gin that has this metadata so I can submit a patch.

[bendichter]

I am not aware of anyone using gates and triggers but let's throw a warning indicating that the starting time of each segment will be inaccurate

[h-mayorquin]

Someone asking about the feature on the SpikeGLX issue tracker:

billkarsh/SpikeGLX#11

(presumably indicates some minority uses the feature)

[h-mayorquin]

Jennifer Colonel also mentioned that she has trained many labs (unspecified) that use these feature.

While fixing some metadata issues on neuroconv I realize that not even the t_start of the multi-probe is correctly set. The fix in #1609 should take care of all the cases.