wrf_dfi_0613

This tar file is based on June 2 WRF repository code (SVN head 1821). The
backward clock should be working in this code, but I will test it again.

Included in this tar file is the diagnostic calculations only. I changed a few
things to accommodate using dpsdt with other diagnostics, corrected the problem 
the Ju-Won found, calculating domain averaged divergence (instead of time change 
of divergence), and removed dmudt (the code was just commented out) since it is 
not computed correctly.

There is still a few remaining issues:

1. surface evaporation was not computed as it is, when using Noah scheme. It is
   not too hard to fix.
2. I would like to check the backward integration a bit more.
3. Will work Jimy to see if we can solve the rain rate issue - this is with
   explicit rain only.

What I did check is the parallel run. 1 proc vs 8 run give close enough results
in the Jan 2000 case. This was done on IBM. It also compiles and runs on Mac.

Wei