Fix bug in drop columns in normalize_coops_stations

[JackReevesEyre-NOAA]

This PR fixes an issue with how columns are dropped in normalize_coops_stations.

As it was, df.drop was called with columns that might not exist, which would raise an exception. There was already logic in place to check if these columns existed, so I moved the df.drop to be within those if statements.

I ran the test suite (make test) and all the COOPS tests passed (overall, 166 passed, 30 warnings).

[pmav99]

Just for the record, another way to tackle this is to use:

df = df.drop([col1, col2, col3], errors="ignore")

https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.drop.html

[JackReevesEyre-NOAA]

The test that fails is only a single version and is related to NDBC tests, not CO-OPS (which is what I changed). So I'm not sure what to do with this... Any ideas @SorooshMani-NOAA @pmav99 ?

[pmav99]

the tests we have some providers are flaky. The first thing you should do is to restart/retry the failing tests, although it doesn't always help. If the failing test is not related to the changes in the code then we may merge even with failures.