Fix adjustment.interpolate_to_reference

[r-pascua]

The interpolation routine itself is fine, but there are two pieces of metadata that aren't updated appropriately for the general case. In particular, the integration_time array is only reshaped, and the channel_width attribute isn't changed. Both of these values should be updated to agree with their new values post-interpolation.

Additionally, in the future we'll need to think about how to adapt this to work with BDA data...