Mtkahypar

This release adds our latest unconstrained refinement algorithm, in addition to other improvements.

Improves compilation times.
Code updates for the deterministic partitioner.

- Remove dependency to KaHyPar
- Exception Handling
- Bug fixes

- Better naming conventions for our different configurations
- Fixed vertex support

New features:
- Interface for implementing new objective function (without having to modify the internal implementation of the refinement algorithms)
- Support for sum-of-external-degree metric
- Mt-KaHyPar can map a (hyper)graph H onto a target graph G now. The objective is to minimize the weight of all Steiner trees induced by the hyperedges of H on G. This objective function is especially useful when modeling wire-lengths in VLSI design or communication costs in distributed system when some processors do not communicate with each other directly or with different speed.

New features:
- Windows build (supports MinGW compiler, but not MSVC)
- Add configuration for partitioning (hyper)graphs into a large number of blocks (e.g., k > 1024).
- Mt-KaHyPar can now optimize the cut metric
- Separate library interfaces for graph and hypergraph partitioning are unified in one library interface (C and Python)

- Mt-KaHyPar is now compatible with the newest version of TBB.
- Mt-KaHyPar can be build from a release archive now (using the `build.sh` script).