Pull request #1062: CAS-14814
Merge in CASA/casa6 from CAS-14814 to master
* commit 'cb65c9b710137b44a5624dadade29b814c6a2a3d':
Remove dead code
labeling, pruning and blob-size counting in memory, avoid temporary copies
Optimize Automasking region labelling and pruning
Main bottleneck is binaryDilation computation. This part of the code has been accelerated now, by: 1.) Remove IPosition constructions from inner loop 2.) Use direct storage pointers 3.) Precompute structure offsets 4.) Return changed from BinaryDilationCore 5.) Remove the full image check in BinaryDilation() 6.) Avoid extra full-plane scans
Optimize Automasking region labelling and pruning
Replace IPosition based region traversal with flat-index DFS
rewrite blob-wise counting using direct image access
prune small regions in a single image pass
