Potential optimization of LLVMExecutableModel::getReactionRates()

[evilnose]

The previous optimizations in LLVMExecutableModel::getReactionRates() attempt to not re-calculate reaction rates when they are updated. However, they fail to account for assignmentRules of boundary variables and would not re-calculate when in fact some variables have changed (see #610). Additionally the bug was present in reaction rates with only time as a variable. Therefore these optimizations are turned off for now (see commented out section in getReactionRates()).

A potential future optimization that is more robust, is to keep track if each variable is constant, and only re-calculate if a reaction has non-constant variables. Additionally, since a reaction rate may depend on another reaction, a table may be needed to keep track which reaction rates have been deemed constant. This should only be done once before each simulation.