fix(netwatch): drain AF_ROUTE socket to avoid macOS route monitor stall

[cbenhagen]

Description

Fixes a read stall in the BSD/macOS route monitor.

netwatch/src/netmon/bsd.rs reads the AF_ROUTE socket through a tokio::net::UnixStream with one AsyncRead::read per await, reading a single message per wakeup and never draining. On BSD/macOS the fd is registered edge-triggered, so leaving data queued can drop the next readiness notification: the reader then stalls with data still queued in the socket (FIONREAD > 0, poll reports it readable) and stops delivering NetworkMessage::Change. The loop only recreates the socket on a read error, and the stall surfaces no error, so it does not recover.

This drain-until-WouldBlock requirement is the documented contract of mio: mio's Poll says an operation "must be performed repeatedly until it returns WouldBlock", else "there is no guarantee that another readiness event will be delivered, even if further data is received".

Changes:

• Read the socket via readable() + try_read() drained to WouldBlock on each wakeup, instead of one AsyncRead::read per await. The UnixStream and its construction are unchanged.
• Keep the existing buffer growth, parse_rib filtering, and socket recreation with exponential backoff.
• Add a deterministic test that read_available drains all queued messages in one readiness episode (via an AF_UNIX datagram socketpair).

Verified on macOS: under sustained route churn the old one-read-per-await reader stalls with data queued and the OS still reporting its socket readable, while the drain-to-WouldBlock reader keeps its socket empty.

Breaking Changes

None.

Notes & open questions

None.

Change checklist

• Self-review.
• Documentation updates following the style guide, if relevant.
• Tests if relevant.
• All breaking changes documented. (None.)

[flub]

Nice, this is looking good! Mostly some nits, thanks for finding this bug!

[cbenhagen]

Thanks for the quick review! Addressed all your comments.

[flub]

Thanks!

(And please use the small "re-request review" button on github, it makes notifications a bit easier to handle on the reviewer-side. It is hidden under the "reviewers" menu to the right of each reviewer and only shows up when you have un-reviewed changed - not the best UX!)