* commiting last version compatible with the PPP-Code indexes built for the PPP-Code indexes (Fall 2013); the VoronoiLeaf nodes use "int objectCount" which is WRONG for M-Index multi-buckets (because of parallel processing of inserts and thread synchronization)
 * next version of M-Index should use "AtomicInteger objectCount" (it is commented out in this version)

   1) in case of multi-buckets, multi-threads, split could have been initiated on a node already split by a different thread
   2) objectCount was transient int, but should have been AtomicInteger
 * fixed approx search in case of multi-bucket when all objects from the multi-bucket are processed (much more objects were accessed than specified by the approx configuration)

 * minor modifs

 * the approx search now can let the PPP-tree cell to revise its distance from query - MinQueryDistanceRevisor
 * fixed a problem with precise range search on disk-oriented M-Index

 * modified query-PPP distance calculators. The main is now WSDCTight, the original one is WSDCalculator, but it was proven wrong (does not fulfill the condition that d(q,parent) >= d(q,child)
 * the "Tight" calculator uses information from an internal node about it's children

 * size of IDs (max number of objects) can be now set in the config file
 * internal cell does not store children in a Map but in an array with nulls
NOT TESTED PROPERLY YET

 * the PPPCOde navigation processors now do not use the SortedCollection at all but they use the std java.util PriorityQUeue and SynchronizedQUeue
 * the orig processors renamed
 * two unncessary classes moved from messif to mindex.private

 * distributed M-Index moved to separate GIT Project
 * modifs due to devel of dm-index

    - ID-object index (to be stored on SSD) is created by two interfaces and a class (package mindex.pppdata)
    - Refinement processor created to refine the candidates within a separate thread

 * minor modifs

 * fixed a BIG BUG in "reorganizeOverfilledBucket" that caused loss of data in case of no sub-index defined and the lowest bucket were overfilled
 * efficiency improvements in PartiallyOrderedCollection
 * much more efficient reading of PPPCOdeObjects from byte[]
 * automatic leaf size in PPPCodeAlg is now calculated so that the resulting capacities are smaller and shrinking with level
 * added richString() in PPPCodeAlgorithm
 * many smaller improvements

 * reading sizes of nodes and data
 

 * added sequential processing of PPP Approx KNN
 * not the whole queue is sorted

 * 

 * setting of QueryPPPDistCal possible via parameter
 * several minor modifications before refactoring (to store only last pivot index in cells in most cases)

* range implemented but SinglePPPCodeDistance calculators

   - to be reworked with the aid of MetricIndex method