Fstd2nc

* Loosen horizontal coordinate matching criteria for series data.
* Blacklist numpy 1.15.3, which had a bug that caused a fatal crash.

* Fixed the --ignore-typvar option, which was crashing.
* Fixed the encoding of depths shallower than 1m. These were
getting truncated to 0m.
* Fixed a bug in handling lists of values in attributes. This was
causing problems when using --rpnstd-metadata on files with a
"!!" vertical coordinate.
* Added two new command-line arguments --diag-as-model-level and
--ignore-diag-level, to deal with diagnostic-level data in the
FSTD file.
* Detect mixture of diagnostic / model levels and inform the user
of how to resolve this using the above commands.
* Package should now be fully compatible with Python 3.
* Added docstring to the fstd2nc.Buffer class.

* Fixed .to_pygeode() method, which was failing on xarray > 0.10.0.
* Added locks to thread-unsafe parts of the code.
* Fixed a float64 upcasting that was done on some record parameters
when using pandas routines.
* Handle whitespace in list-based command-line arguments, e.g.
--vars="VAR1, VAR2, VAR3".
* Fixed a crash when using --rpnstd-metadata on vertical levels.

* Added 'station' as a coordinate for 'Y' series data.
* Some incremental speedups in constructing Buffer objects.
* Command-line option to rename variables.
* Command-line option to create an 'ensemble' axis over varying
etikets.
* Some new features / improvements geared towards using this package
in Python scripts:
* Read FST records in their natural order in the xarray interface.
* Added an optional 'header_dict' parameter to Buffer, to allow it to
remember the headers of files that have been previously scanned.
* Added a to_iris() method to generate an Iris CubeList object.
* Added a to_pygeode() method to generate a PyGeode Dataset object.

* Re-enabled unlimited time dimension for applicable files.
* Renamed zaxis,zeta,eta,sigma axes to just 'level'.
* Removed old anciliary_variables stuff for A/B vertical coeficients.
* Added formula_terms for vertical coordinates, along with a "pref"
(reference pressure) scalar variable.
* Loosened criteria for matching vertical (!!) record when there's only
one vertical record in the file.
* Fixed bug when there are multiple reftimes in a file (issue 15).
* Added support for time-dependant lat/lon coordinates from LA/LO
records.
* New --exclude option to remove variables and degenerate axes from
the netCDF file.
* New --profile-momentum-vars and --profile-thermodynamic-vars options
to specify which variables are on which type of level.
Note: only needed for profile data.
* New --missing-bottom-profile-level option to work around a bug in
profile data coming from rpnphy output.
* Fixed a bug in the times for profile / series data.

* New --rpnstd-metadata-list to select a subset of RPN metadata
attributes to copy to the netCDF file.
* Made --minimal-metadata the default. The old behaviour (copying all
RPN metadata) can be triggered with a new --rpnstd-metadata option.
* Coordinate-related RPN metadata is no longer copied, since the values
are already encoded in the coordinate variable.
* Restored support for 'A', 'B', and 'G' grids.
* Fixed the repeated longitude in global grids - value is now 360
instead of wrapping back to 0.
* Use consistent lat/lon/station coordinates for series data when using
a --filter on the ip3 values.
* Unused axes are now pruned from the file. Mostly affected series
files, which had an extra time axis from the unused datev values.
* Vertical axes are now encoded as floats, since that's the precision
of the original ip1 values. The pandas version of the iterator was
upconverting these to doubles, which would cause different results
depending on whether pandas was installed or not.