* Global precomputed distance cache implemented as an precomputed distances filter (for fast substitution in existing implementations)
* Incompatiblity: PrecomputedDistancesFilterFactory requires a local abstract object as an argument to allow also automatic filter registration.
* LocalAbstractObject.getDistanceFilter registers the new filter and overrides any existing one (it did not overwrite it before)
* JavaDoc fixes

KMeans pivot chooser fix in reselecting seeds on invalid input data (selecting the same object as multiple seeds)
New helper method in OperationStatistics

* Updated k-means/k-medoids pivot chooser (still the issue with object on hyperplane boundaries to solve)
* CoveragePivotChooser fixes
* New algorithm for recall/precision evaluation of various queries
* New algorithm implementing a simple text-based protocol (used in master thesis by Jakub Hanko)

OperationTime query parameter is set on completing a query all the time (regardless the global stats are enabled/disabled)

* Fix in deserializing bucket id object key

* Classes of ObjectFloatVectorNeuralNetwork move from the private repository

* CoreApp extension by .repeatEvery (time specification)

* KMeans pivots chooser and hierarchical chooser

* New dump structure operation
* KNN operation fixes the number of objects added (upon refining its answer)
* KMeans pivots chooser and hierarchical chooser

* Updated java compiler parameters in POM
* Added support for shutdown hook registration

Master

Approved-by: David Novak <novak.david@gmail.com>

minor convenience updates; float vector's method toString now prints the raw data (needed for GetObjectOperation#answer called by CoreApplication)

Batch ranking query

major update that adds an UpdateLog that is automatically replayed in case of a hard system fall

Alow construction of BatchKNNQueryOperation using any Iterator,
not only StreamGenericAbstractObjectIterator

Master