Basic running working

[richfitz]

This PR just throws together enough of the idea of basic model running so that we can sensibly discuss the next features. Notably missing but coming in the next few PRs:

• More than one block of particles (we'll do this sooner rather than later)
• Interesting time handling, allowing steps of size dt (where 1/dt is an integer)
• Error handling
• Parallelism
• Comparison with data, resampling
• Generation of the wrapper code (contents of src/walk.cpp, which can be generated easily)
• Design and generation of an interface for the R side (all the dust2_walk functions)

Also the current version just uses the walk model which is special because it has 1d state, so we'll need to improve handling of state passing back and forth by putting some dimensions on input and output

[weshinsley]

OK - Think I get all of this!