Make `priority` a computed property of effects

Currently all the effects have to statically sorted by their class in EffectOrdering, which leads to an artificial proliferation of overlapping effect classes like Substitution, Insertion, Deletion, ComplexSubstitution. Really, all of those should be specific run-time instantiations of a single Substitution effect. To give single residue substitutions a lower priority then e.g. deletions, we should have a dynamic effect priority like:

def priority(self):
    flags = [
        self.changes_protein_length(),
        self.predictable_coding_effect()   
        self.overlaps_transcript()
        self.overlaps_gene(),
    ]
    total = 0
    for i, flag in enumerate(flags):
        total += 2 ** i * flag
    return total

Start by reading EffectOrdering and the existing Substitution, Insertion, Deletion, and ComplexSubstitution effects. Trace how their class-based ordering is applied, then verify that runtime priority produces the intended ordering for effects with different protein-length, coding, transcript, and gene impacts. Done means the overlapping effect classes can be represented by a single Substitution effect without losing ordering behavior.