In the (rare) case that one has multiple shells with different number of orbitals (e.g. p-d-systems) the old code crashed when converting GFs from the second shell as matrix dimensions were not fitting (the tmp matrix was always created for the first shell).
merged unstable into py3
* include the major changes of a94f8ed94280ed8bfa7c46b8f811aa402d7c8981
of blockstructure.py
* adapted all code to python 3
* fixed all tests
* moved the plovasp C++ code to c++/triqs_dft_tools/converters/vasp
* added global header triqs_dft_tools/triqs_dft_tools.hpp
* python dir based on single cmakelist file
* registered C++ tests for plovasp
* corrected imports for py3 tests for plovasp
* corrected block order in sigma_from_file and srvo3_Gloc
* exchanged ref files for sigma_from_file, srvo3_Gloc, SrVO3.ref.h5
* moved vasp converter bash scripts from dir shells to bin dir