pf: postpone clearing of struct pf_pdesc

[kp]

Summary

Postpone zeroing out pd until after the PFI_IFLAG_SKIP/M_SKIP_FIREWALL
checks. We don't need it until then, and it saves us a few CPU cycles in
some cases.
This isn't expected to make a measurable performance change though.

Pointed out by: markj
Sponsored by: Rubicon Communications, LLC ("Netgate")

[mjg]

The thing to do in the long run is to stop zeroing it upfront to begin with. KMSAN, if operational, can be used to validate there are no uninitialized uses. I tried to do it few weeks back but ran into bugs.