Support more types for DLD dataframe extra columns

[philsmt]

Building on #166, this PR relaxes the type requirements a bit to work with both pulse- and train-resolved data in pd.Series, xr.DataArray or directly KeyData objects. Naturally the pulse index must be compatible, i.e. use the same first two columns.

[takluyver]

Is it a deliberate choice to make this method public? I think it will also appear in the docs, and I tend to think that keeping the documented API relatively small, so people can easily see the methods they want, is a good thing. But YMMV. 🙂

[philsmt]

It kind of is. I don't expect wide usage of it for now, but it seemed useful to not restrict aligning columns to the loading code if it's actually independent of it.

[takluyver]

The implementation looks good to me.

It doesn't seem 100% ideal to be making APIs that look like load_some_data(with=some_other_data), as opposed to some form of load_some_data() + some_other_data. But I think that's hard to avoid given the constraints of what you're doing here, so go ahead.

[philsmt]

Thanks for review!

Indeed I still want a PulsePattern.label_data-esque API that can take data in as many different formats as possible and align it. But it seemed unlikely to get this right quickly, so I opted for a smaller integration in the DLD component itself. It should be easy to continue supporting this API on top of a more generic one later.

[takluyver]

In that case, LGTM