add_MARBL: introduce _USE_MARBL_TRACERS CPP macro

[mnlevy1981]

#22 did the following:

1. Remove MARBL from git submodules
2. Introduce MARBL into Externals_MOM.cfg

This issue will continue the work:

3. Wrap the bulk of MARBL_tracers.F90 and all calls into the module from MOM_tracer_control_flow.F90 in #ifdef USE_MARBL_TRACERS
4. Make sure MOM builds without _USE_MARBL_TRACERS
5. Update CESM build to include always include CPP

[mnlevy1981]

An interesting thought from Matt L: if the CPP is available, we should use it in the CESM build. Meaning compsets that are not set up to run with MARBL should not build MARBL.

1. This effectively makes MARBL a build-time specification despite tracer count being run-time (but how often do CESM users decide to turn on a tracer package by hand rather than via compset?)
2. What will be the mechanism for turning MARBL on by default via compset? It seems like MOM%MARBL in the compset long name should set an env_build.xml variable - does MOM still use OCN_TRACER_MODULES or do we need something new like MOM_ENABLE_MARBL?

[mnlevy1981]

Given how this work is breaking down, I think #21 (comment) should be addressed in a separate issue once this work is complete

[mnlevy1981]

Having trouble triggering the abort when running with USE_MARBL_TRACERS = True and not building with -D_USE_MARBL_TRACERS, need to track call to call_tracer_flux_init() [that's where the generic tracer abort is found]

[mnlevy1981]

It turns out that call_tracer_register() is a better place for the error check (may eventually end up in both routines?). With some CIME mods to add the -D_USE_MARBL_TRACERS cpp, this issue is resolved