This is needed to cleanly read the time stamps without having to drop
anything. Having an enum value that is only used in one schema does not
disturb the others, right?

Fix read-in of a NetCDF dates list

See merge request !17

I don't think it can work the way it was (by constructing a string from
the address of a single char, no length), at least not for arbitrary
files... don't think it can hurt to put the length we already know
explicitly?

Including using regex for increased robustness

The problem came when the command / argument index gets high in the input data
editing. For example, an editing command 8 as well as an editing command 880. In
this case, the arguments to command 880 where delivered to command 8 (matching
the common "*::ARG8"). In order to prevent this partial matching, the pattern
has been extended with the necessary final "::".

The ILWR parametrization plot has been extended with the Carmona All Sky parametrization. The same plot has been done for all supported Clear Sky parametrizations and the documentation has been extended.

And used the opportunity to make code more c++11 like

Now if the DEM is not needed for any station, it won't be loaded at all. This
means it is now safe to provide a DEM "just in case" a station has not yet been
pre-computed

The separation between low level curl / json parsing and the higher level logic
is now much better done. This also cleanly prevents exposing curl / picoJson to
the rest of MeteoIO through includes.

Now the proper casting from string to double/int is checked as well as for sampling_rate that must be >0

The logic has been changed, somes data structures have been removed
while some others were expanded. The API had changed since the early
version of this plugin and this needed to be accounted for. Lots of
unnecessary variables have been removed, the code relies on more c+11
constructs. The documentation has been updated.

But there is still lots left to do, there are some inconsistencies...

This should always be done in data generators or in Snowpack itself

Moreover, some loops have been re-written in c++11 for clarity (when applicable) and the meteo parameter name extraction from config keys has been rewritten using regex (more readable and more robust).

The SPDX identifier is been added back

Finally, the filename parsing is now also done with regex which should help with long term maintennace... Also some code cleanup.

The documentation has also been updated (thanks Paul Billecocq for reporting
this!). WINSTRAL_LISTON will be updated in another commit...

Moreover the naming has been changed, referring to "synthesizers" instead of
"synthGenerators" for clarity and consistency