Mobility fitting compatibility and context futures

[nury12n]

This adds the following:

• Compatiblity with Mobility fitting materialsgenomefoundation/kawin#19. Only change here is for the thermochemical equilibrium residual to ignore any datasets with TRACER or DIFF in the output field. Then the residual functions implemented in kawin will provide support for tracer and interdiffusivity residual functions
• Option to scatter the context to the Dask workers before emcee is called rather than moving it to the workers during each MCMC iteration. This helps with memory management where pickling/unpickling of the context should only occur once before emcee and anytime a worker is restarted.
  • This is currently set to be the default behavior
  • Some log statements are also added to track the total memory and memory per worker at each iteration

[bocklund]

Is there any reason why we wouldn't want to always use futures to scatter the context (and eliminate the option entirely)?

[bocklund]

Sorry, two small changes that occurred to me. Otherwise LGTM.