Making the Fortran codebase more maintainable

[marcadella]

I noted the following things that I believe will make it difficult to maintain the codebase in the future:

• Mixing of attributes from different models in a same top-level derived type class (ex: vegn_tile_type, cohort_type). Instead, attributes that are specific to a certain model should be grouped in one derived type which can be added to the top-level type as one attribute. Separation of concerns in important!
• Mixing of attributes from different time resolutions (ex: annualGpp, dailyGpp). Instead, it would be much better to group all daily (resp. annual) diagnostics in their own derived type and add a pointer to cohort_type. The advantage is that one can discard the old diagnostic data by dropping the pointer and replacing it with a new fresh instance (with appropriate default values). Subroutines such as Zero_diagnostics, which are highly undesirable (!!) would then become unnecessary.
• Non initialization of all attributes. Instead any attribute defined in a derived type should be initialized with a sensible default value or a dummy value (ex: -9999, or better: IEEE NaN), if no good default exists. In other words, any attribute defined in datatypes.mod.f90 and classdefs.mod.f90 should have a default value.
• A detail, but there is no reason to add .mod in the Fortran file names. This extension is used for compiled files and it is a bit confusing to have it in the source files.
• The analysis directory should not be committed to master.