Refactor `interpolate` to work on pd.Series

[danielhuppmann]

The interpolate() method currently works on a wide representation of the timeseries data, and then casts the result back to the long-form _data series. It would probably be more efficient to work on a group-by of the long-format series in the first place, eliminating the need to cast to wide and back.

Also, the method raises an unexpected error if not all timeseries have a datapoint before and after the year argument, i.e., the following call for df.interpolate(2007) fails...

					2005	2010
model	scenario	region	variable	unit
model_a	scen_a	World	Primary Energy	EJ/yr	3.0	6.0
			Primary Energy|Coal	EJ/yr	0.5	NaN