I'm not sure if that would be possible, as that requires to figure out raw memory address of script_execute at runtime (assuming that it's a function at all) while having exactly zero information about it's structure or whereabouts.
However, it is not required to check for queued errors at the end of the step - you can check whenever, such as immediately after you do something that might have caused an error. Some people even made "try" / "catch" macros for GMS2 to store/check the current number of queued errors