Establish philosophy of relation of metaprob to clojure

One way to define metaprob is as the language that the meta-circular interpreter interprets. An additional constraint comes from the desire to make traces the only compound data structure, with a complete inductive theory (e.g., the ability to copy a trace using advertised trace primitives without any loss). Clojure is very different; it has all sorts of things the meta-circular interpreter won't understand, and it has an open ended set of opaque data types not amenable to induction. Yet, a desire to harmonize the two to the extent possible has been expressed. Also expressed has been a desire to add encapsulated data to metaprob (via protocols). It would be good to have an articulated common philosophy giving direction to the development effort, so as to avoid having everyone go off in different directions.

Raised by Desiree, and suggested by mention of protocols in #77. Related: #59, #5, #52, #8. Assiging to VKM (for now at least) because I think ultimately only he can make decisions about this.