Timestamp discontinuity with `CVSource` class

[h-mayorquin]

The condition to fallback to system time in the CVSource class is the following:

https://github.com/rjdmoore/fictrac/blob/9ac055e52d89f49f492a8eb4e1f7c5b8cd6df40a/src/CVSource.cpp#L162-L165

But I think the condition should be strict, that is, I think the condition should be _timestamp < 0. Otherwise, timestamps which are 0 are ignored and the fallback of the system time is used instead which creates a time discontinuity. As an example, this is the data that I got by running the algorithm in the sample.mp4:

	frame_counter	sequence_counter	movement_direction	movement_speed	timestamp	alt_timestamp	delta_timestamp
0	0	0	-0	0	1.69019e+12	4.14554e+07	0
1	1	1	5.79849	0.0246007	33.3333	4.14554e+07	-1.69019e+12
2	2	2	3.5273	0.0211707	66.6667	4.14554e+07	33.3333

See the large discontinuity caused by the first timestamp being 0.

To confirm that the first timestamp is indeed 0 we can run ffprobe:

ffprobe -i sample.mp4 -select_streams v:0 -show_entries frame=best_effort_timestamp_time,pkt_pts_time,pkt_dts_time,pkt_duration_time -of default=noprint_wrappers=1 -v quiet | head -n 20
pkt_pts_time=0.000000
pkt_dts_time=0.000000
best_effort_timestamp_time=0.000000
pkt_duration_time=0.033333
pkt_pts_time=0.033333
pkt_dts_time=0.033333
best_effort_timestamp_time=0.033333
pkt_duration_time=0.033333
pkt_pts_time=0.066667
pkt_dts_time=0.066667
best_effort_timestamp_time=0.066667
pkt_duration_time=0.033333
pkt_pts_time=0.100000
pkt_dts_time=0.100000
best_effort_timestamp_time=0.100000
pkt_duration_time=0.033333
pkt_pts_time=0.133333
pkt_dts_time=0.133333
best_effort_timestamp_time=0.133333
pkt_duration_time=0.033333