ArrayRangeToLoop might incorrectly assume full range

[nmnobre]

When processing ECMWF's NEMO code, I came across an issue in trc_oce.f90. Indeed, it'd seem that:

prgb(:,:) = zrgb(2:4,:)

is being transformed to and replaced by:

!$omp parallel default(shared), private(idx,idx_1)
!$omp do schedule(static)
do idx = LBOUND(prgb, 2), UBOUND(prgb, 2), 1
  do idx_1 = LBOUND(prgb, 1), UBOUND(prgb, 1), 1
    prgb(idx_1,idx) = zrgb(idx_1,idx)
  enddo
enddo
!$omp end do
!$omp end parallel

[LonelyCat124]

I think this is a duplicate of issues mentioned in #1841 (second comment)

[sergisiso]

I am fixing this in #1841