Add from_pandas and from_xarray

[SimonHeybrock]

We have previously added from_dict and to_dict, but it has proven to be of limited use, since it is too specific to scipp's data schemas. For better interaction with other common libraries in the Python ecosystem, we should add from_pandas and from_xarray.

to_pandas and to_xarray may be added as a second step, but they are more tricky since there may be scipp features that cannot be represented in those libraries. So let us start with:

from_xarray:

• Support DataArray and Dataset
• data maps to values
• can we handle units? xarray does not support specifics, so unless we want to rely on xarray attributes with specific names we have the set them as dimensionless in scipp
• Note that xarray attributes have no direct equivalent in scipp. Attributes of xarray data arrays or dataset may be stored as scalar variables in scipp attributes.
• xarray distinguishes coord with and without index. Coords with index should map to scipp coords, coords without index to attributes.

from_pandas:

• Not too familiar with this, need to check if there is a way to identify columns as coords

Overall, it is probably best to start simple. Some of the details above (such as attr or unit handling) can be done later, no need to get everything working in the first pull-request.

See also:

• Discussion in .convenience functions converting to/from xarray #1849
• xarray's equivalent of from_pandas http://xarray.pydata.org/en/stable/generated/xarray.Dataset.from_dataframe.html

[SimonHeybrock]

Instead of implementing from_pandas, we might choose to simply use from_xarray(xr.Dataset.from_dataframe). Downside is that the user would need to install xarray, but if it avoids a lot of code it may be the best choice for now.