Preconditioning in weak-constraint data assimilation algorithms

[arango]

Various updates to the weak-constraint 4DVar data assimilation algorithms:

• Andy implemented the Ritz preconditioning in congrad.F which is used by the weak-constraint 4DVAR algorithms (W4DVAR and W4DPSAS).
• The weak-constraint impulse forcing was reformulated. A new module frc_weak.F is introduced.
• The routine impulse.F will be obsolete soon when we finish the new W4DPSAS algorithm . The impulses are now processed in new routine wrt_impulse.F.
• Added the adjoint relaxation terms ad_uv2drelax.F and ad_t3drelax.F which were missing.
• A new cpp option was introduced DATALESS_LOOPS which can be used to test the convergence of the Picard iterations of the representer model in W4DVAR.

Many thanks to Andy Moore for his help in coding and testing all these new upgrades.