Add chemical gradient following

[mountaindust]

A really robust solution would accept an initial chemical concentration and use the fluid velocity plus specified boundary conditions to solve for future times. However, solving the equations involved with a robustness for different boundary conditions will take a lot of structure, and with immersed boundaries it's not clear it would be correct even then. So for the time being, just provide a place to load time-varying data for a chemical concentration.