Standardized missing values from CCPP framework to use in CCPP physics

[climbfuji]

The CCPP framework should provide missing values that can be used by the CCPP physics.

We will need missing values for every type and kind, i.e. integer(4), integer8, real4, real8, and all kinds defined by CCPP, too: real_kind_phys, ...

Thoughts/Ideas:

• use standard names such as ccpp_missing_value_real_kind_phys, ccpp_missing_value_real_(star_)4, ... ?
• should the framework dictate the values or can the host model overwrite them?
• for integers huge() is the only reasonable value - possibly for others, too
• the unit conversion tools and other auto-converters need to ignore such values, i.e. should use where (a /= missing_value_for_type_and_kind_of_a) ... endwhere