JDK-8177044 introduced the _scan_top array containing knowledge about what part of the region should be scanned for references during GC.

This information can not only be used for refinement (update rs), but also during scan rs phase.

Further, it proved useful to rearrange the filters in the card scanning code a bit to
- not do work if we can quickly determine we won't use it
- optimize a bit for duplicate card scans by moving the card table check first