Object key is set to BucketInfoObjectKey when kNN query is evaluated to remember the time when the object was discovered and added to the answer.

It uses new MESSIF operation - PartitionedBulkInsertOperation

* Log messages about which edge in spanning tree was used

* During the split a half of longest edges in spanning tree are tested for a balanced split.

* Ordering of edges in minimum spanning tree (for finding an even split) is inverted (from longest edge to shortes)
* Added log reports (to stderr) about uneven split