Fix LOR in if_lagg

[kbowling]

Summary

What happens is, when setting up an LACP lagg, we come through lacp_linkstate earlier via pr_addport, and it checks the interface's media status. We can't hold an rmlock while doing that with i.e. cxgbe

Example LOR:

begin_synchronized_op with the following non-sleepable locks held:
exclusive rm if_lagg rmlock (if_lagg rmlock) r = 0 (0xfffff80186601a08) locked @ /d0/kev/freebsd/sys/net/if_lagg.c:703
stack backtrace:
#0 0xffffffff8097d1d0 at witness_debugger+0x70
#1 0xffffffff8097e5fe at witness_warn+0x45e
#2 0xffffffff8047d53c at begin_synchronized_op+0x3c
#3 0xffffffff80488021 at cxgbe_media_status+0x31
#4 0xffffffff80a1d540 at ifmedia_ioctl+0x170
#5 0xffffffff804873c6 at cxgbe_ioctl+0x276
#6 0xffffffff8242542c at lacp_linkstate+0x4c
#7 0xffffffff824259e1 at lacp_port_create+0x1f1
#8 0xffffffff82422b90 at lagg_ioctl+0x1340
#9 0xffffffff80a13a03 at ifioctl+0xe43
#10 0xffffffff80982db4 at kern_ioctl+0x2c4
#11 0xffffffff80982a6e at sys_ioctl+0x16e
#12 0xffffffff80d25fc9 at amd64_syscall+0x549
#13 0xffffffff80d0715b at Xfast_syscall+0xfb

Test Plan

Set up an LACP lagg

[mav]

It won't work this way, since lagg_port_destroy() in case of error require WLOCK to be locked. If go this way, it at least should be relocked before calling lagg_port_destroy().

[kbowling]

In D11711#242651, @mav wrote:

It won't work this way, since lagg_port_destroy() in case of error require WLOCK to be locked. If go this way, it at least should be relocked before calling lagg_port_destroy().

That makes sense, updated

[mav]

I have no objections.