The concurrent cleanup and secondary free list handling has several disadvantages:
- adds another lock in the LAB allocation path
- additional concurrent phase

Measurements show that freeing regions is very fast (sub-ms) for a significant set of regions. Particularly with JDK-8180415, where the very expensive remembered set scrubbing will be removed, the additional time freeing regions takes is negligible (actually even without that change).

Remove concurrent cleanup completely, the secondary free list and everything related.