Otherwise we compare non-zero with zero values.

- Parameters with more than one component are split into components for
  comparison
- The absoulte criteria gets scaled with the maximum absolute value in
  the parameter set
- It is possible to pass a zero threshold, i.e. the threshold under which
  values get treated as zero, for each parameter or
  parameter component as dict
- More verbose output when verbose
- Adapt the tests

It got ignored previously.

While the previous implementation might have fixed some previous problem,
it failed at completely making the DotDict ordered. Added a test that
failed with python3 without this test.

Btw. Steffen told me the reason for this pyhton2/3 difference: Python
had security issues with non-randomness in hash generation and fixed
that in python3.

dune-grid is only suggested.

Feature/separate scripts

Separating the scripts from the modules and installing them through setuptools.

See merge request !48

Why not?

Merge branch 'feature/separate-scripts' of ssh://conan2.iwr.uni-heidelberg.de:20022/quality/dune-testtools into feature/separate-scripts

Conflicts:
	python/scripts/analysis.py
	python/scripts/has_static_section.py

gridconstruction: Support ALU, do not catch errors (prevents silent unexpected behaviour)

See merge request !45

Feature/dev package

This fixes #76 .

See merge request !47

make && make test is now sufficient to run the tests.

1) HAVE_MPI comes from config.h
2) The MPI flags need to be added to the created targets