Fix duplicate timestamps in merging InSAR with GPS

[scottstanie]

The full GPS workflow was sometimes creating duplicate InSAR rows:

It was not obvious in the graphs/linear fits, but I was manually editing these out for other summary statistics.

The problem was multiple GPS rows (from adjacent days or different times on the same day) were matching to the same InSAR epoch when using merge_asof with tolerance="1D", direction="nearest"

Now, after the merge, we keep only the closest GPS row per InSAR epoch by:

1. Preserving the matched InSAR epoch as a column (insar_time)
2. Computing the time difference between GPS and InSAR times
3. Selecting the GPS row with the minimum time difference per InSAR epoch