witness: report a thread that can't sleep in WITNESS_WARN

[glebius]

Summary

Although this is not related to locking, but seems to be the right
place to catch violators, given that WITNESS_WARN is already used
widely in the kernel.

[glebius]

Immediately caught in bridge(4):

NET_EPOCH_ENTER(et);

callout_drain(&sc->sc_brcallout);

[kib]

sys/kern/subr_witness.c

2140

IMO it is cleaner to inline THREAD_CAN_SLEEP() because the whole function operates on td.
The right hand of && is not style-compliant.

[markj]

sys/kern/subr_witness.c

2144

[pouria]

In D55316#1265083, @glebius wrote:

Immediately caught in bridge(4):

NET_EPOCH_ENTER(et);

callout_drain(&sc->sc_brcallout);

Are we waiting for this?
Do you want me to work on bridge?

[glebius]

• Apply Kostik's & Mark's suggestions