Prevent empty seg in random_segment which causes tests to fail stochastically

[zm711]

Basically we have stochastic failures in this test

python-neo/neo/test/coretest/test_segment.py

Lines 46 to 66 in 74c97fd

	def test_times(self):
	for seg in self.segments:
	# calculate target values for t_start and t_stop
	t_starts, t_stops = [], []
	for children in [seg.analogsignals, seg.epochs, seg.events, seg.irregularlysampledsignals, seg.spiketrains]:
	for child in children:
	if hasattr(child, "t_start"):
	t_starts.append(child.t_start)
	if hasattr(child, "t_stop"):
	t_stops.append(child.t_stop)
	if hasattr(child, "time"):
	t_starts.append(child.time)
	t_stops.append(child.time)
	if hasattr(child, "times"):
	t_starts.append(child.times[0])
	t_stops.append(child.times[-1])
	targ_t_start = min(t_starts)
	targ_t_stop = max(t_stops)
	self.assertEqual(seg.t_start, targ_t_start)
	self.assertEqual(seg.t_stop, targ_t_stop)

because currently we sometimes generate a completely empty segment with nothing in it so the list of t_starts is empty which causes the min function to fail. This makes us have to re-run the test CI to get a result. Instead I made the function recursive such that each segment needs to have at least one of the types of children objects for testing purposes (to be clear it just needs one so if the n_child is 0 for all types of children then it fails which is what causes the test to stochastically fail).

The other option would be to change the test to accept the fact we have an empty segment. So let me know what you think @apdavison !

[apdavison]

I think this is a good solution.