[Feature suggestion] xarray support

[westurner]

An xarray backend for einops.

References:

• AbstractBackend:
  

  einops/einops/_backends.py

  Line 62 in 1bee724

  class AbstractBackend:

• pytorch named tensors <> einops integration #39 (comment)

[westurner]

(Here from xarray-contrib/pint-xarray#3 (comment) re: namedtensors , though originally the question was about uncertainties support and +- in sympy, and whether (mu +/- sigma) is actually just PDF/PMF) sympy/sympy#5305 (comment)

[westurner]

• Use Case: GenCast reads netCDF weather data with xarray, and wants to implement einops abstractions for portability

[arogozhnikov]

very poorly written suggestion.

https://github.com/arviz-devs/xarray-einstats exists as a separate package, it takes into account that xarray tensors already have names associated with dimensions, not something einops deals with