Provide a mean for users to perform waveform extraction

[GaelleChapuis]

For some applications, such as Petrina or Han who want to get single spike waveforms, they may need to extract waveforms from the raw data themselves.
There are 3 components to this:

1. Know which processing to perform on the raw data before extracting the waveform
2. How to find the channels on the raw data associated with the spikes
3. How to best save the data for efficient memory usage

On point 2:
In order to know on which channel a spike is, right now the information is hard to get: only the cluster holds the channel information, not each spike.

To get the information, one would need to apply the inverse of the drift correction

[oliche]

See ibl-neuropixel 1.4.0