ragged.regular_to_ragged to accept xarray.dataarray or xarray.dataset input

[selipot]

Right now ragged.regular_to_ragged accepts only numpy array. I think we should expand that function or write a new function that would automatically convert a regular data xarray.dataarray/dataset to a ragged xarray.dataaray/dataset.

[selipot]

Such a function would do something like the following automatically with a number of checks. Assuming the regular xarray input dataset is ds_in with dimensions traj and dimension coordinate obs. And after having applied a mask to all the variables with dimensions (trajectory,obs).

n_traj = ds_in.sizes["traj"]
n_obs = ds_in["lat"].count().values

# Initialize the Dataset
ds_out = xr.Dataset(
    coords={
        'id': ('traj', ds2["id"].data),
        'obs': ('obs', range(n_obs))
    }
)
for var in ds_in.data_vars:
    dum, rowsize = regular_to_ragged(ds2[var].to_numpy(), fill_value=np.nan)
    ds_out[var] = xr.DataArray(dum, dims=("obs"), name=var)
ds_out["rowsize"] = xr.DataArray(rowsize, dims=("traj"), name="rowsize")