skbuff changes (iwlwifi, rtw88, rtw89)

Hi,

I've had these chanes sit in my tree for a while incrementally done during
the entire LinuxKPI malloc fixes over the last 8-ish(?) months.
I refactored them into a few coherent commits this evening and pushed them
to main.

Noticeable changes -- in theory none (1).

In practise for now we'll do two allocations rather than one per packet
again as the skb (header) comes from a UMA zone and data comes from LinuxKPI
__kmalloc (or if a tunable is set still from contigmalloc).

As one of the commit messages indicates this is to work towards not having
to copy the data between skb and mbuf anymore in future versions, at
least in the RX path.  TX will be a bit more funny to deal with but one
step at a time.

In case anyone notices any real problems please let me know sooner than
later.

Lots of health,
Bjoern

(1) one mentioned in the commit message, if you are running a latop with
more than 64GB of memory and an older 9xxx or earlier Intel card with
iwlwifi, you may have to set the compat.linuxkpi.skb.mem_limit=2 (not 1
like for rtw88) tunable in loader.conf.  That said you probably would
have had that problem in the past already.

-- 
Bjoern A. Zeeb                                                     r15:7