as this class is often needed it is now part of the mesa_pd core

To break unneeded dependencies in postprocessing the SQLite part is
moved to its own module and namespace.
Therefore in the include and the function calls
"postprocessing" hast to be removed or replaced by "sqlite" e.g.:
#include "postprocessing/sqlite/SQLite.h" ->
#include "sqlite/SQLite.h"

postprocessing::store... -> sqlite::store...

there is no need for a special rotation matrix class

BlockForestDomain is now passed via argument rather than
recreated in every call.

moved initialization of BlockForestDomain into refresh function
This way the BlockForestDomain can be refreshed (e.g. as part of
the load balancing) without recreating it.

This was lost in the migration from boost to std as boost::hash<T> (and thus boost::unordered_map<T,V>) uses hash_value<T> if available and std::hash<T> (and thus std::unordered_map<T,V>) does not.