A User should be able to modify the scoring function

[mtbarta]

A user should be able to evaluate different retrieval scoring methods and ranking functions.

This enables faster benchmarking and ad hoc experimentation of new scoring functions so that retrieval can be updated over time.

This could be as simple as modifying parameters in bm25, or as complex as a learned hyperplane to apply.

---

There needs to be a set of primitives that enable Scorers to be built. Additionally, documents need to be scored within the context of the query -- we stopped short of this to launch the query engine. A document should be the composite score of all of its nodes.

Also note, that we need to disconnect fast fields/ context nodes from scoring so that we can achieve better multithreading.