Finalize droping spikextractors, remove its use in toy_example.py

[h-mayorquin]

On #196 @CodyCBakerPhD did most of the hard work of removing the spikextractors as a dependency and pushed it as an optional work flow.

What is left is to eliminate this completely. In the toy example spike-extractor is used to generate a set of frames in which the spikes occur to transform them to roi (that is, not the spikes times but the actual frames, given a sampling frequency, in which the spikes occurred). We should be able to reproduce this functionality here and eliminate the dependency for good.

So, where is this in the code, the actual functions from spikeextractors are here:

roiextractors/src/roiextractors/example_datasets/toy_example.py

Lines 126 to 131 in b71f17d

	rec, sort = se.example_datasets.toy_example(
	duration=duration,
	K=num_rois,
	num_channels=1,
	sampling_frequency=sampling_frequency,
	)

A recording and a spikextractor. They only use is here:

roiextractors/src/roiextractors/example_datasets/toy_example.py

Lines 140 to 151 in b71f17d

	raw = np.zeros((len(sort.get_unit_ids()), rec.get_num_frames()))
	deconvolved = np.zeros((len(sort.get_unit_ids()), rec.get_num_frames()))
	neuropil = noise_std * np.random.randn(len(sort.get_unit_ids()), rec.get_num_frames())
	frames = rec.get_num_frames()
	for u_i, unit in enumerate(sort.get_unit_ids()):
	for s in sort.get_unit_spike_train(unit):
	if s < rec.get_num_frames():
	if s + len(resp) < frames:
	raw[u_i, s : s + len(resp)] += resp
	else:
	raw[u_i, s:] = resp[: frames - s]
	deconvolved[u_i, s] = 1

Note that the the recording extractor (rec in the code) is only used to get the sampling frequency and the number of frames which is not really necessary. I will re-structure this in a small PR to make this obvious. The only difficult thing to remove is the production of the spike trains in line 144:

roiextractors/src/roiextractors/example_datasets/toy_example.py

Line 144 in b71f17d

for u_i, unit in enumerate(sort.get_unit_ids()):

So, a function that produces frames in which spikes occur in some reasonable fashion should suffice for this small toy example.

[h-mayorquin]

In #205 I added indications of what needs to be done for when someone has time to take this on:

roiextractors/src/roiextractors/example_datasets/toy_example.py

Lines 140 to 152 in 5b74133

	num_frames = rec.get_num_frames() # TODO This should be changed to sampling_frequency x duration
	num_of_units = sort.get_unit_ids() # TODO This to be changed by num_rois
	# convolve response with ROIs
	raw = np.zeros(num_of_units, num_frames) # TODO Change to new standard formating with time in first axis
	deconvolved = np.zeros(num_of_units, num_frames) # TODO Change to new standard formating with time in first axis
	neuropil = noise_std * np.random.randn(
	num_of_units, num_frames
	) # TODO Change to new standard formating with time in first axis
	frames = num_frames
	for u_i, unit in range(num_of_units):
	unit = u_i + 1 # spikeextractor toy example has unit ids starting at 1
	for s in sort.get_unit_spike_train(unit): # TODO build a local function that generates frames with spikes