Our previous HCCPARAMS1 encoding was actually advertising 4 streams,
as it's expressed in power-of-2 terms. In ppci_xhci_init_ep, we'd
get it wrong by not really validating MAXPSTREAMS and treating it as
the number of streams rather than the exponent for a power-of-2. That
isn't great, but at worst we may over-allocate rather than access things
out-of-bounds.
Sponsored by: The FreeBSD Foundation (aokblast)