Improve docs for use of new Units table to store spike times.

[tjd2002]

There are several places in the documentation where we still recommend users use the (potentially deprecated) 'Clustering' processing module to store spike times.

After we decide what the correct recommendations are regarding use of Units, we should fix all these:

• https://pynwb.readthedocs.io/en/latest/tutorials/general/file.html#processing-modules Mentions Clustering
• https://pynwb.readthedocs.io/en/latest/tutorials/domain/ecephys.html#extracellular-electrophysiology-data Talks about 'storing spike data', but doesn't mentions Units
  https://pynwb.readthedocs.io/en/latest/overview_nwbfile.html#processing-modules. Doesn't mention Units, does refer to Clustering

And possibly more...

A useful resource are the release notes on the refactoring that led to the creation of Units:
https://github.com/NeurodataWithoutBorders/nwb-schema/blob/dev/docs/format/source/format_release_notes.rst#improved-storage-of-unit-based-data

@tjd2002 and @angevineMiller will take this on

There are a bunch of related issues on this:
NeurodataWithoutBorders/nwb-schema#194 (comment)
NeurodataWithoutBorders/nwb-schema#239
^ @bendichter

• Decide whether Clustering is deprecated, or should be kept around for specialized use cases like Streaming.

[tjd2002]

Clustering and ClusterWaveforms is also used extensively in the tutorial for creating an extension:

pynwb/docs/source/extensions.rst
https://pynwb.readthedocs.io/en/stable/extensions.html?highlight=Clustering

It's awkward/bad if we use a deprecated type in these examples. For now, I think we should yank it out, then replace it with another type later. Since this is just a generic placeholder for a DataInterface, I have replaced it with 'SpikeEventSeries' or 'EventDetection'.

[tjd2002]

If we decide to formally deprecate Clustering for 2.0, then #797 should be merged. If not, then we should just cherrypick some of the doc fixes in there, to at least make sure we don't steer people towards Clustering.