Test coverage

[atspaeth]

There are a lot of code paths that aren't covered in the unit tests. Here's a list of the ones I've noticed:

• The SpikeData.from_thresholding() constructor and its use of butter_filter().
• SpikeData.__getitem__ calling .subtime() and .subset() in ways that make sense (this would be a bit of an API demo too!).
• Any simple test for SpikeData.append() and .concatenate_spike_data()
• Generate synthetic data to test SpikeData.isi_skewness(), .isi_log_histogram(), and .isi_threshold_cma() (which should use and therefore test cumulative_moving_average()). Also burst_detection().
• Tests outside the happy path in SpikeData.latencies().
• A test which mainly serves to document the invariants of SpikeData.randomized() (such as preserving neuron_attributes).
• The FR smoothing algorithm of population_firing_rate().