Fixes to normalization.F and def_info.F

[arango]

There is a bug when computing the 4DVAR background error covariance normalization factors for the adjustment of surface tracer flux. The code around line 2707 of normalization.F needs to be modified to use HnormSTF instead. Many thanks to Gregoire Broquet for reporting this bug.

Also fixed def_info.F to write the curvilinear rotation angle in polar coordinates applications. The variable angler was only written to output NetCDF files when spherical=.TRUE. This is kind of problematic in polar coordinates applications. This was changed to write angler when CURVGRID is activated. Many thanks to Dale Haidvogel for bringing this to my attention.