[flang] Fix overflow detection for folding SUM/PRODUCT

[klausler]

Summary

The overflow detection code in the templates that fold SUM and PRODUCT
was checking for overflow before performing the reduction, not after.
Fix and add tests.

[PeteSteinfeld]

All builds and tests correctly and looks good.

flang/lib/Evaluate/fold-reduction.h

         auto prod{element.Multiply(array->At(at))};
         overflow |= prod.flags.test(RealFlag::Overflow);
         element = prod.value;
       }
     }};
+    auto result{Expr<T>{DoReduction<T>(*array, dim, identity, accumulator)}};
     if (overflow) {
       context.messages().Say(
           "PRODUCT() of %s data overflowed"_warn_en_US, T::AsFortran());
-    } else {
-      return Expr<T>{DoReduction<T>(*array, dim, identity, accumulator)};
     }
+    return result;
   }
   return Expr<T>{std::move(ref)};
 }
 
 // SUM

                          .value.Subtract(next.value, rounding)
                          .value;
         element = sum.value;
       }
     }};
+    auto result{Expr<T>{DoReduction<T>(*array, dim, identity, accumulator)}};
     if (overflow) {
       context.messages().Say(
           "SUM() of %s data overflowed"_warn_en_US, T::AsFortran());
-    } else {
-      return Expr<T>{DoReduction<T>(*array, dim, identity, accumulator)};
     }
+    return result;
   }
   return Expr<T>{std::move(ref)};
 }
 
 } // namespace Fortran::evaluate

flang/test/Evaluate/errors01.f90

     real, parameter :: bad2 = abs(cmplx(huge(0.),huge(0.)))
     !CHECK: warning: DIM intrinsic folding overflow
     real, parameter :: bad3 = dim(huge(1.),-.5*huge(1.))
     !CHECK: warning: DIM intrinsic folding overflow
     integer, parameter :: bad4 = dim(huge(1),-1)
+    !CHECK: warning: HYPOT intrinsic folding overflow
+    real, parameter :: bad5 = hypot(huge(0.), huge(0.))
     !CHECK: warning: overflow on REAL(8) to REAL(4) conversion
     x = 1.D40
   end subroutine
 end module