Tweaks to allow CRTM_2.2.3 to compile on Fujitsu (#214)

TYPE: no impact

KEYWORDS: WRFDA, CRTM 2.2.3, Fujitsu compiler bug

SOURCE: Fujitsu, Japan

DESCRIPTION OF CHANGES:

Statements
rts_stats(:,1) = rts_stats(:,1)/factor
and
rts_stats(1:n_channels,1) = rts_stats(1:n_channels,1)/factor
do not compile on Fujitsu due to their compiler bugs.
Change it to explicit loops
DO l = 1, n_channels
rts_stats(l,1) = rts_stats(l,1)/factor
END DO

This part of code is in a function that is not used anywhere else.

LIST OF MODIFIED FILES:
M var/external/crtm_2.2.3/libsrc/CRTM_RTSolution_Define.f90

TESTS CONDUCTED:
1. crtm_2.2.3 can compile on Fujitsu with the modified code.
2. WRFDA regtests

var/external/crtm_2.2.3/libsrc/CRTM_RTSolution_Define.f90

@@ -641,7 +641,10 @@ FUNCTION CRTM_RTSolution_Statistics(rts, rts_stats) RESULT( err_stat )
         rts_stats(l,1) = rts_stats(l,1) + rts(l,m)
       END DO
     END DO
-    rts_stats(:,1) = rts_stats(:,1)/factor
+    !rts_stats(:,1) = rts_stats(:,1)/factor
+    DO l = 1, n_channels
+      rts_stats(l,1) = rts_stats(l,1)/factor
+    END DO
 
 
     ! Compute the standard deviation
@@ -650,7 +653,10 @@ FUNCTION CRTM_RTSolution_Statistics(rts, rts_stats) RESULT( err_stat )
         rts_stats(l,2) = rts_stats(l,2) + (rts(l,m) - rts_stats(l,1))**2
       END DO
     END DO
-    rts_stats(:,2) = SQRT(rts_stats(:,2)/factor)
+    !rts_stats(:,2) = SQRT(rts_stats(:,2)/factor)
+    DO l = 1, n_channels
+      rts_stats(l,2) = SQRT(rts_stats(l,2)/factor)
+    END DO
 
 
     ! Replace the algorithm identifier