Bugfix/multi material with internal variables

[btalamini]

This PR comes from a request from a collaborator to be able to do multi-material simulations with history-dependent materials. This overlap of capabilities had been intentionally omitted due to time constraints. The PR does two things:

1. Implement missing pieces for multi-material solid mechanics with materials with internal state. I made an implementation decision that bears noting: I padded the internal variable array such that it has enough storage for the largest set of internal variables. That is, the internal variable array has shape (ne, neqp, nint), where ne is the number of elements (same as before), neqp is the number of quadrature points per element (same as before), and nint is the maximum of the number of internal variables for a material over all materials. This wastes memory, but makes the implementation easier. We can revisit this later if necessary. Probably the entire multi-material design needs to be rethought and unified with the single material implementation. I noted another potential performance concern, which is that the elements in each block are not enforced to be consecutive. This is a bigger concern, since it will likely be a big speed hit. I opened issue Ensure elements in blocks are consecutive #25 to keep track of this.
2. Write a tutorial in a Jupyter notebook for multi-material simulations. This is the first real documentation of any kind in optimism. It's meant to help my collaborators use the library. It's the first attempt to address Documentation #20 .

I didn't add unit testing of the multi-material capability, so I opened issue #26 . As I noted above, I think the design needs to be re-thought anyway. I would like to get opinions from Tupek and others on this.