makeAntPco and distinct antenna element bands

[ourairquality]

Describe the bug

The antenna PCO interpolation for missing frequency entries is an interesting feature. Looking at the code it appears to choose the two closest frequencies and use these for interpolation.

Taking a look at the PCO entries of an example antenna in igs20.atx, TRM105000.10 and plotting the PCO vs wavelength gives the plot below. The slope of the variation within a frequency band is opposed to the step across bands, perhaps due to a stacked antenna element design. If there are at least two calibration entries in each band (as for TRM105000.10) then this would not be a problem for this algorithm as two close entries would be found from the same band. But for a low end antenna there might be limited calibration entries, perhaps just L1 and L2, and in this case interpolating using entries in different bands might be far worse than just using the single closest entry.

So might makeAntPco be usefully modified to only choose close entries within the same band, within a limited frequency range, and if only one entry is found then to not give up but to just use this one close entry.

Also noted a possible related feature in antPco() (auto_fill_pco) that appears to fallback to GPS L2 if a PCO entry is not found. At a guess this it for the missing GPS satellite L5 entries, to fallback to the L1/L2 entry. But might this mask the interpolation path and cause it to use L2 even if there were an entry close to L1 (sorry not clear how that code works but it appears to be related, and it was not clear if these are intended to be mutually exclusive options).