SpikeInterface objects and xarray DataTree conversion functions

Hi,

I use xarray datatypes for most of my scripts and I feel all spikeinterfaces objects could fit in the xarray datatree datatype with almost no drawback (we keep lazyness, chunks, ...). I would really enjoy to be able to use spikeinterface objects and functions for most of my pipeline, but every now and then, have a function to convert to a xarray datatree and vice-versa (perhaps requiring a specific structure of the datatree).

Are conversions functions from and to xarray datatree a feature that may be added to SpikeInterface in the future?

In a more broad function, my understanding of spikeinterface machinery is that it has its own chunk/lazyness handler, and those mostly operate on time chunks. Would there be any drawback in using xarray with Dask as back-end instead, except for the complete code overhaul ?

The issue names no files, tests, or entry points. Start by reviewing SpikeInterface's existing object and function interfaces alongside the proposed xarray DataTree structure; define the required conversion directions and compatibility boundary before deciding whether a Dask-backed overhaul is viable.