And update .pre-commit-config.yaml

As "extends" stopped working for whatever reason.

Fixes issue #4.

This is a merge.

It seems that the previously used method to shorten string station IDs
would not produce unique IDs in some cases.

Summary of changes:
    * Lint code with "isort", "flake8(helled)" and pylint
    * Add option "--selected-stations-fpath" to csv2obsoul command
    * Fix issue in csv2obsoul command when station IDs are integers
    * Other minor bugfixes

Changed:
    * Rename the "--selected-stations" option in the csv2obsoul
         command to "--selected-stations-fpath"
    * Make invalid selected_stations_fpath raise error

Fixed:
    * Crash in --selected-stations-fpath opt
    * Crash in csv2obsoul command
    * Crash in clustering & select when no station gets accepted
        in the clustering.

This is a merge commit.

Instead of just issuing a warning. Otherwise users may not notice that
the stations have not ben filtered until they examine the output
files.

Change the "--selected-stations" option in the csv2obsoul command to
"--selected-stations-fpath". This is to make it more obvious that this
refers to a file path.

The old syntax using "--selected-stations" should still work though, as
argparse tries to match the options from passed abbreviations.

Caused by a reference to select_stations before the variable had been
defined.

To simplify the code and fix an unpacking error in the csv2obsoul
command.

They already get shortened when reading data in read_netatmo_csv (which
is used by all other file reading routines so far in netatmoqc).

This should solve the issua at FMI with failure for integer station IDs.
See issue <#3>.

This is a merge commit.
The previous commit in the branch was 22e5a889

And fix an issue with retrieval of DTG metadata in save_df_as_obsoul.