Documentation improvements for 3.0.0: Auto-reduction solver and State_Het

[jimmielin]

In reference to previous conversation in #50

I've added two commits here to

1. describe rosenbrock_autoreduce in the Numerical Methods section, and
2. add a blurb about using derived type objects to store state information for KPP use, as used by GEOS-Chem in State_Het. Also added the use of commonIncludeVars.H to avoid cluttering up .kpp code. Please feel free to edit 😃

[RolfSander]

The autoreduce description looks good.

The additional text for F90_GLOBAL is very interesting, but it is also
very GEOS-Chem specific. I think the main points here are that F90_GLOBAL
can also be used to:

• define derived-type objects

and

• #include additional input files

Maybe the description in the KPP manual could focus on these general
points and refer to the GEOS-Chem manual for specifics about State_Het
and commonIncludeVars.H?

[yantosca]

I also think the autoreduce section looks good.

As @RolfSander said, we can remove the GEOS-Chem-specific reference but say that you can add derived a derived-type definition in F90_Global. I can work on that.

[yantosca]

I updated the F90_RATES and F90_RCONST sections.

[RolfSander]

Thanks, @yantosca!

I added that USE must come before any variable definitions.