Implementation of T1920 and meaningfull error message if a user dont has mpi4py installed

Summary:
there are still some problems running e.g. 'test_till_random_material.py', I think we need this 'Valgrind debug'.
I just want to share the code with you in that way.

commented the ProjectioError to stop segmentation fault

Merge branch 'master' of ssh://c4science.ch/source/muSpectre into feat/ProjectionError_T1904

Implementation of a ProjectionError for even grids, need discussion

there are still some problems running e.g. 'test_till_random_material.py', I think we need this 'Valgrind debug'.
I just want to share the code with you in that way.

commented the ProjectioError to stop segmentation fault

Merge branch 'master' of ssh://c4science.ch/source/muSpectre into feat/ProjectionError_T1904

Merge branch 'feat/ProjectionError_T1904'

Merge branch 'feat/ProjectionError_T1904_solved'

WIP: there are still problems in material_linear_elastic4.cc and.hh

WIP: small changes, still not running

ENH: Added scalar push_back function for scalar fields.

BUG: InternalVariables needs to be a tuple that holds two scalar maps.

Small cleanup of MaterialLinearElastic4 interface

• evaluate_stress and evaluate_stress_tangent took rvalue references instead of lvalue references to the internal variables
• the #define header guard of material_linear_elastic4.hh was identical to the one in material_linear_elastic3.hh, resulting in MaterialLinearElastic4 not being visible in the python binding files.

modified CMakeLists to prevent segfault with gcc5

done like it already was done for https://c4science.ch/D168

WIP: still errors in python_material_linear_elastic4_test.py,

the stress is computed wrong.

WIP: I only missed to stage two files

Small fixes on materiallinearelastic4

Material was returning a T4MatMap instead of T4Mat, thus returning
a reference to a temporary. also, debugging print statements were in
the wrong method.

Found and fixed bug in MaterialLinearElastic4

The material computes the first Lamé constant λ and shear modulus µ in
add_pixel. Then in compute_stress(_tangent), it is called with λ,
µ (as can be checked by looking into the order in which the fields are
arranged in the constructor), but the methods' declarations were
written expecting Young's E and Poisson's ratio ν, so they wrongly
recomputed λ(E,ν) and µ(E,ν) which then were totally off.

Cleand version: docs corrected, sdt::cout's deleted, all tests run

--> should be ready for arc diff.

Merge branch 'master' of ssh://c4science.ch/source/muSpectre

Merge branch 'feat/store_Lame_consts_per_pixle_T1920'

ENH: corrected error message if one has mpi4py not installed.

Test Plan:
have some discussion and fix the Segmantation fault problem with 'test_till_random_material.py'

have some discussion and fix the Segmantation fault problem with 'test_till_random_material.py'

Reviewers: junge

Differential Revision: https://c4science.ch/D178