Current state of checkpointing in Stan

[mtwest]

Just noticed this thread on output serialization future options, which seems to have grown out of some
thoughts from @rok_cesnovar partly about restarting failed sampling runs.