Support spikeinterface

[JoeZiminski]

(currently draft for discsussion)
This PR will add spikeinterface as a data backend.

The idea would be for viewephys to take either a path to a .bin file, data array or a dictionary of spikeinterface recordings where the keys are the preprocessing step. e.g.

viewer = EphysBinViewer(   # will rename this or have multiple interfaces
    {"raw": raw_rec, "filtered": filtered_rec, "cmr": "cmr_rec}
)

then the checkboxes will be auto-populated with the keys and the data will be extracted from the relevant recording.

We can have a single SpikeInterfaceDataModel that holds this dict and exposes all of the requirements of the EphysBinViewer (see #30). The dict of recordings could live on the SpikeInterfaceDataModel, we can do my_recs[step].get_traces(...) and have some logic that every recording in the dict shares the same sampling rate, geometry etc.

WDTY?

[JoeZiminski]

Hey @oliche @rai-pranav this PR is not yet finished (requires tests and some further investigation into the memmapping) but I won't finish it until next week so this could be a good time for an interim review