Change default `xarray_open_kwargs` to use `chunks="auto"`

[dougiesquire]

Is your feature request related to a problem? Please describe.
Currently, by default, Intake-ESM opens files using chunks={}.
For netCDF files:

• with xarray < v2023.09.0, this gives one chunk per file, which is sometimes a bad choice;
• with xarray >= v2023.09.0, this gives the file chunking, which is often a bad choice.

Describe the solution you'd like
Default to chunks="auto" to use dask auto chunking. This should give better chunking in most cases.

Happy to submit a PR.

Describe alternatives you've considered
Passing the kwarg every time I use Intake-ESM

[dcherian]

gives the file chunking

Isn't this a multiple of the file chunking?

[dougiesquire]

I think chunks="auto" gives multiples of the file chunks, but chunks={} gives the file chunks.

[mgrover1]

This sounds reasonable to me... feel free to go ahead and open a PR!

[dougiesquire]

It was decided to leave the default as chunks={} since chunks="auto" does not work with object dtype (which is common).

[charles-turner-1]

It was decided to leave the default as chunks={} since chunks="auto" does not work with object dtype (which is common).

We've had a chat about this this morning & I think we can probably cleanly handle this case - gonna reopen the issue & take a look.

[charles-turner-1]

• Add chunks='auto' support for cftime datasets pydata/xarray#10527
  Following the linked PR above being merged, we should be able to change this default without worrying (for up to date versions of xarray.