Fix lval extension for compressed assignment in constant function.

The type of extension (zero/sign) is determined by the expression
type, not the operand type, so we need to cast to the expression
type before extending the value.

(cherry picked from commit 241b6723e5)

net_func_eval.cc

@@ -252,8 +252,8 @@ void NetAssign::eval_func_lval_op_(const LineInfo&loc,
 	    break;
           default:
 	      // The left operand must be cast to the expression type/size
-	    lv = cast_to_width(lv, rv.len());
 	    lv.has_sign(rv.has_sign());
+	    lv = cast_to_width(lv, rv.len());
       }
       switch (op_) {
 	  case '+':