Support SciPy sparse matrices for alpharank code, implement Stochasically Stable Distributions

[l-ouis]

Evolutionary dynamic Markov chains are sparse matrices, so using SciPy sparse matrices and operations can save memory.

Also implements SSD: https://arxiv.org/pdf/1207.1424, which is an exact algorithm that computes the stochastically stable distribution of a Markov chain (and can be used similarly as alpharank)

• move Kosaraju.py into ssd.py and fix formatting
• support sparse matrices in the SSD reduce step

@Eric-Ewing

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[lanctot]

Thanks @l-ouis ! Can you sign the CLA?

[Eric-Ewing]

We should catch the case when meta_strategy_method is None (or not provided) and make it default to alpharank to ensure backwards compatability

[lanctot]

Can you fix the indentation to be consistent? (two-space indent)

E.g kosaraju.py appears to use 4-space indent but there might be others.|

Edit: original_ssd.py as well. Or maybe you don't intend to keep that one in the PR? (i.e. is it the original implementation that you're fitting into OpenSpiel, for instance)