Add configuration for removing entities or descendants of entities from lexicon generation

[cthoyt]

Need some notion of exclude lists in the lexica construction to remove frequently occurring, but not specific entities like "Disease" and "Role"

So like we want to say that "these top level terms are irrelevant" or "this subtree is irrelevant" in addition to/after the subsetting operation gets applied

Part of the solution is in #8 for a simple exclusion list