Formula rules: no need for timestep if delta_a doesn't appear

[timhutton]

Currently, formula rules such as Experiments/CornusAmmonis/smoke.vti have to declare a timestep parameter even though they write directly into a and don't use delta_a. The generated kernel source includes the forward Euler step even though it isn't used.

Should include forward-Euler for only those chemicals that write into delta_chem.