Contour block SS

[jarlebring]

New attempt for the contour integral method block SS method (previously contour_assu) described in Asakura, Sakurai, Tadano, Ikegami, Kimura, A numerical method for nonlinear eigenvalue problems using contour integrals, JSIAM Letters, 2009 Volume 1 Pages 52-55

Example:

julia> nep=SPMF_NEP([[0 1 ; 1 1.0], [1 0 ; 0 0]], [s->one(s),s->exp(1im*s^2)]);
julia> λ,V=contour_block_SS(nep);
julia> norm(compute_Mlincomb(nep,λ[1],V[:,1]))
1.3148059621411757e-15

There were some details missing in the original JSIAM-paper, so I had to use techniques from other papers.

Still not quite finished: Cleanup needed. We could also add other quadrature rules and ideally do code sharing with beyn, but it requires quite a bit of reordering of the code.

• Change beyn quadrature to use Juliaesque types (and make it extendible)
• quadrature code sharing with beyn
• feature: integration over ellipse in block_SS
• cleanup: function comments
• cleanup: make warnings like in beyn
• tutorial on user-specified quadrature rule

[jarlebring]

This new way of specifying quadrature might break use of beyncontour for some people (e.g. work of @wrs28?) but it allows a user to write their own integrator and it has a lot of code sharing btw block_SS and beyn.

[jarlebring]

The unification of handling of error messages requires some more thought process and deserves a different PR.

[jarlebring]

I'm now feeling more or less ready with this PR. Several recent CI-failure seems to be due to a change of test computers. I think travis now ran on a computer with less memory and therefore failed for the WEP-tests. Unrelated to this PR.