Ensure parser produces an indepotent output

Reordered fields in the spec should not produce any diff. I don't think this is the case today, need to verify. `parseIR()` should ensure schemas are processed in the same order, and there's likely another step needed for individual fields within each schema.