in the earlier commit the changes to the saturation vapor pressure
modules were inadvertently negelected.  These come with this commit
along with some clean up of the directory structure.

`

Revise handling of package metadata and build-system

See merge request !4

To define generic functions we use the closure construct which
introduces a way to make a function with the given parameters so that
when it is called it is executed with those parameters without having to
pass the parameters.  This allows the function calls for different
quantities to share a common interface, and more allows for a cleaner
separation between a function and its parameters, the latter being used
to construct the function.

https://black.readthedocs.io/en/stable/guides/using_black_with_other_tools.html

For further information see PEP517/518

Ignore python byte code.

restructured treatment of saturation vapor pressure

See merge request !3

Replaced use of es_liq by es_liq_default, likewise for ice to make the
intent clear.

Replaced use of TK and PPa by T and P to ensure consistency.

Released as version 0.4, includes reformatting of analytic formulation
of saturation vapor pressure to better emphasize the commont form of the
calculation for ice and liquid, similar to what is done for teten's
formulas.  Now both it and Teten's serve as 'master' formlae, with liq
and ice being particular implementations.

Attempted to clean up the presentation of the Teten's and Romps formulae
for saturation vapor pressuer, incorporating improved fitting constants
for the former.  These were derived by fitting to Wagner and Pruss for
the temperature range 270 K to 310 K and to Wagner et al for ice over
the range 230K to 260 K.  The fitting is not particular sensitive and
fitting with noice leads to slight changes in the fits, but the new
constants are robustly an improvement over the old ones, which had
inconsistent triple points vapor pressures over ice and liquid. I also
double checked the 'Romp's' formulae, and refrained from calling these
Romps as they are just straightforward integrations of the
Clausius-Clapeyron equation under the Rankine-Kirchoff assumptions.

In this branch additional impelmentations of the saturation vapor
pressure were impelmented.  These different treatments are not meant to
be exhaustive, but the teten's murray formulations were introduced
because these are used by ICON.  The murphy-koop formulation is the
standard for super-cooled liquid water.  The examples ipynb script was
augmented to present the new functionality.

in the atmospheric sciences there are many different moist static
energies.  They differ due to the choice of reference enthalpies for
water phases, which are connected by the phase change enthalpies, and
thus have a single degree of freedom.  This we give to the reference
enthalpy (enthalpy at T0) for vapor.  The choice of hv0 matters because
not all reference quantities can be zero for all phases of water, and so
the choice influences how the reference quantity weights different water
phases.

the function has been defined in a rather expressive way to bring out
this property and make it possible to compute one or the other of the
moist static energies by choosing a different reference state for the
vapor enthalpy.

By passing the function that is to be inverted it is possible to write a
generic inversion function, which we call invert_for_temperature and
invert_for_pressure respectively.  These work for all off the
thermodynamic funcions, i.e., theta_e, theta_l, etc.

I experimented with an invert_for_humidity, but this often did nto
converge so I did not introduce this functionality.

Add CI to run doctests

See merge request m219063/aes_thermo!2

I added the calculation of the moist adiabat from the integration of the
first law to the functions, and a generalized definition of the 'dry'
potential temperature, which can account for moisture effects on
thermodynamic parameters.

An additional example was added to show the added functionality.

the murphy and koop (2005) and hardy (1998) formulations for
saturation vapor pressure are sometimes desired.  The former
for their ability to handle saturation over supercooled liquid,
the latter because this equation is frequenly used for the
processing of radio sonde data

added this helper function to make the effect of qt!=qv more clear, a
situation that is found when air is saturated, or has condensate in
disequilibrium.  This is now used in the definition of the moist
potential temperatures.

- added better referenced and more complete docstreams for all of the
functions.
- modified docstring formatting to better align with standards (google)
- cleaned up some stray cells in the examples notebook

- doctests were corrected and now exist for most functions. `
- repository was restructured so example scripts are not outside of main
  repository

Fixed the call and expected outcome statements of the doctest strings in
the funciton documentation.  In doing noticed that for the mixed
supersaturations we should use the minimum, not maximum values, which
reverts a false correction from a previous commit.

Fix doctests

See merge request m219063/aes_thermo!1

The functions are imported into the top-level namespace directly while
the constants are accessible through their submodule namespace.

Some of the routines assume the existence of a directory to
collect plots.  This adds the directory to the repository.

This is a new release.  In addition to cleaning up existing code,
it adds error checking (which identified a few bugs), and a number
of examples taken from my archived (unreleased) library.  This
includes an analysis of different saturation vapor pressure
formulations, the effect of the saturation vapor pressure formulation
on the potential temperatures, and differences in calculations of ths
lifting condensation level.