File writing for paradiag solution.

[JHopeCollins]

This will write every timestep in the paradiag object passed in to separate files.

The way I added it to the asQ/init.py, the module has to be separately addressed as "asQ.post". I did this on the assumption that it won't always be needed and is somewhat tangential to the actual paradiag method. Is this appropriate or should it be included directly into the top level asQ module like the paradiag and DiagFFTPC classes?

[JHopeCollins]

There are now two functions (with somewhat ambiguous names), do we want to keep both, or just one?

• write_timesteps: each timestep is written to a seperate file, numbered by the timestep. Each time-rank does this for its own timesteps independently of the other ranks
• write_timeseries: all timesteps are written through the same file so appear as a timeseries through the pvd file. This is done by passing the solution at each timestep to the first time-rank, so is essentially serial in time.

The way the MPI calls are written in write_timeseries is possibly overly cautious. I think this could be simplified a bit by just sending everything as non-blocking sends up front, and then having rank 0 just collect them in turn.