cpu_engine: optimize RLE clipping

- Reduce span intersection complexity from O(N×M) to O(N+M) with
  two-pointer traversal.
- Report an empty clipping result as unsuccessful.