We should reject malformed sequences of port policies
C Tor, and our specifications, accept all of these as valid port-based exit policies:
accept 1-500,2-501 reject 100-200,1-10 accept 1-500,500-510
Arti, on the other hand, rejects any port-based exit policy if its ranges are not in order and non-overlapping.
We should clarify the correct behavior (I lean towards reject) in a proposal and do the correct parsing-correctness migration dance.