Fix an error with O31D exceeding vertical dimension kte (#1803)

TYPE: bug fix

KEYWORDS: O3 output, o3input=0, RRTMG-LW

SOURCE: Reported by Samm Elliott, fixed internally

DESCRIPTION OF CHANGES:
Problem:
In 4.4, we introduced ozone output o3rad when o3input = 0, that is to use ozone profile provided by the RRTMG code. But the use of output array accessed memory with the vertical dimension > kte. This can cause segmentation fault on some systems.

Solution:
The array is now limited to use only when vertical dimension is less than kte.

LIST OF MODIFIED FILES: 
M     phys/module_ra_rrtmg_lw.F

TESTS CONDUCTED: 
1. Yes
2. Are the Jenkins tests all passing?

RELEASE NOTE: Fixed a memory access error when filling output array o3rad for o3input = 0. No impact on results, but the error could potentially cause the model to abort.

phys/module_ra_rrtmg_lw.F

@@ -11647,10 +11647,9 @@ SUBROUTINE RRTMG_LWRAD( &
  INTENT(IN ) :: XLAND, &
  XICE, &
  SNOW
-!ccc Added for time-varying trace gases.
+! Added for time-varying trace gases.
  INTEGER, INTENT(IN ) :: yr
  REAL, INTENT(IN ) :: julian
-!ccc
 
 !
 ! Optional
@@ -11962,7 +11961,7 @@ SUBROUTINE RRTMG_LWRAD( &
 ! Pressures are in mb
 !
 
-!ccc Read time-varying trace gases concentrations and interpolate them to run date.
+! Read time-varying trace gases concentrations and interpolate them to run date.
 !
  IF ( GHG_INPUT .EQ. 1 ) THEN 
  CALL read_CAMgases(yr,julian,.false.,"RRTMG",co2,n2o,ch4,cfc11,cfc12)
@@ -11983,8 +11982,6 @@ SUBROUTINE RRTMG_LWRAD( &
  cfc12 = 0.538e-9
  END IF
 
-!ccc
-
 ! latitude loop
  j_loop: do j = jts,jte
 
@@ -12414,7 +12411,7 @@ SUBROUTINE RRTMG_LWRAD( &
  else
  do k = kts, nlayers
  o3vmr(ncol,k) = o3mmr(k) * amdo
- o31d(k) = o3vmr(ncol,k)
+ if (k.le.kte) o31d(k) = o3vmr(ncol,k)
  enddo
  endif