a collection of WRFDA fixes related to ensdim_alpha use cases

[jamiebresch]

TYPE: bug fix

KEYWORDS: WRFDA, ensdim_alpha

SOURCE: Jamie Bresch and Craig Schwartz (NCAR), Yi-Chuan Lo (Central Weather Bureau, Taiwan)

DESCRIPTION OF CHANGES:
The PR includes multiple bug fixes that are somewhat intertwined and related to the use of ensembles under not-usually-used scenarios.
The root cause of the various problems (reported by Craig Schwartz and Yi-Chuan Lo) is the inclusion of alpha control variable in the ill-designed be data structure.
The list of fixes:

1. fix minimization for precondition_cg=true (default is false) and orthonorm_gradient=true (default is false) when ensdim_alpha>0. The fix also helps with offline varbc (related to unallocated be%cv_mz when use_background_errors=false).
2. fix for 3DEns-FGAT (num_fgat_time>0 and ensdim_alpha>0 and je_factor=1.0). The bug was introduced when trying to fix a bug for 4D-EnVar (commit b449e865).
3. fix offline varbc when use_background_errors=false (commit d69d7bb breaks this capability) by still allocating be%cv_mz when use_background_errors=false to avoid memory segmentation fault. A bonus fix is to skip updating analysis and writing out wrfvar_output for offline varbc (VARBC.out is the only meaningful output) to save computing resources.
4. skip reading be.dat when ensdim_alpha>0 and je_factor=1.0 (i.e. full ensemble mode) to avoid the hassle of providing a climatology be.dat that is not actually used in assimilation.

LIST OF MODIFIED FILES:
M var/da/da_main/da_solve.inc
M var/da/da_minimisation/da_calculate_gradj.inc
M var/da/da_minimisation/da_calculate_j.inc
M var/da/da_minimisation/da_minimise_cg.inc
M var/da/da_minimisation/da_transform_vtod_wpec.inc
M var/da/da_minimisation/da_transform_vtod_wpec_adj.inc
M var/da/da_minimisation/da_transform_vtoy.inc
M var/da/da_minimisation/da_transform_vtoy_adj.inc
M var/da/da_setup_structures/da_setup_be_regional.inc
M var/da/da_vtox_transforms/da_transform_vpatox.inc
M var/da/da_vtox_transforms/da_transform_vpatox_adj.inc
M var/da/da_vtox_transforms/da_transform_vptox.inc
M var/da/da_vtox_transforms/da_transform_vptox_adj.inc
M var/da/da_vtox_transforms/da_transform_vtox.inc
M var/da/da_vtox_transforms/da_transform_vtox_adj.inc
M var/da/da_vtox_transforms/da_vtox_transforms.f90

TESTS CONDUCTED:

1. the fixed code (implemented in different versions of WRFDA) has been used extensively in different applications
2. WRFDA regression test would be needed, but I didn't do it
3. Jenkins tests all passed

RELEASE NOTE: WRFDA has been fixed and enhanced for applications related to offline varbc (use_background_errors=false), 3DEns-FGAT (num_fgat_time>0 and ensdim_alpha>0 and je_factor=1.0), precondition_cg=true and orthonorm_gradient=true when ensdim_alpha>0, full-ensemble mode (ensdim_alpha>0 and je_factor=1.0).

[jamiebresch]

Please find result of the WRF regression test cases in the attachment. This build is for Commit ID: 8cb7a42, requested by: jamiebresch for PR: https://github.com/scala-computing/WRF/pull/1603. For any query please send e-mail to David Gill.

Test Type              | Expected  | Received |  Failed
= = = = = = = = = = = = = = = = = = = = = = = =  = = = =
Number of Tests        : 23           24
Number of Builds       : 60           58
Number of Simulations  : 159           155        0
Number of Comparisons  : 96           93        0

Failed Simulations are: 
None
Which comparisons are not bit-for-bit: 
None

[liujake]

Will look into this early this afternoon.

[liujake]

Looks good. WRFDA regression tests passed.

[liujake]

@davegill I approved 4 PRs by Jamie. You may give another approval when you have a chance.