Need cleaner way of specifying global tutor dependent constraints #63

So on the AL_HTML side I'm starting to work on they stylus tutor bridge which is would be the first non-CTAT tutor so far. I've had to change the global rule constraints a bit to fit this new tutor. These are things like the selection should be editable and the arguments should not be empty. These constraints are usually written inside the ModularAgent with a certain set of object attributes in mind, (i.e. there needs to be a 'value' or 'contentEditable' attribute for the CTAT constraints). In the interest of making the ModularAgent tutoring system agnostic we should probably move these outside of Modular agent, and have them defined in a similar way as operators where custom ones can be written outside of the core code. Additionally it might make sense on the AL_HTML side to have Tutor specific defaults for certain arguments which are implied simply by choosing a particular tutor, but could be overridden in the training.json