Expanded NL_BULK_FLUXES to other algorithms

[arango]

Expanded the usage of NL_BULK_FLUXES as reported in previous update ​src:ticket:368 on 10/06/09. Recall that now NL_BULK_FLUXES is used to handle the reading of surface fluxes computed by the nonlinear model (bulk_flux.F or a basic state NetCDF file) in the adjoint-based algorithms.

The routines tl_get_data.F, tl_set_data, ad_get_data.F, and ad_set_data.F follow the same strategy as in the representer model routines (rp_get_data.F and rp_set_data.F).

This means that the functionality of BULK_FLUXES and NL_BULK_FLUXES can be used in all the adjoint-based algorithms including all the 4D-Var drivers.

Removed the obsolete IOM C-preprocessing option. This cleaned some of the CPP directives in the code.