ofz+ubsan: limit double range to sal_Int32 limits

wmf fuzzing

Change-Id: I37b437717f064c6c85cd383315adf4e989486412
Reviewed-on: https://gerrit.libreoffice.org/43272Tested-by: Jenkins <ci@libreoffice.org>
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>

include/basegfx/numeric/ftools.hxx

@@ -22,7 +22,7 @@
 
 #include <rtl/math.hxx>
 #include <basegfx/basegfxdllapi.h>
-
+#include <limits>
 
 // standard PI defines from solar.h, but we do not want to link against tools
 
@@ -59,6 +59,10 @@ namespace basegfx
     */
     inline sal_Int32 fround( double fVal )
     {
+        if (fVal >= std::numeric_limits<sal_Int32>::max())
+            return std::numeric_limits<sal_Int32>::max();
+        else if (fVal <= std::numeric_limits<sal_Int32>::min())
+            return std::numeric_limits<sal_Int32>::min();
         return fVal > 0.0 ? static_cast<sal_Int32>( fVal + .5 ) : -static_cast<sal_Int32>( -fVal + .5 );
     }