[RTI-15742] Close connection after writing the last checkpoint

[abijr]

Close connection after writing the last checkpoint when receiving a terminate event.

This PR adds a state_timeout (see gen_statem state-timeout), this timeout will be cancelled if there's a state change (which will happen if we get a response). If we want to trigger the timeout regardless of state changes, we can use a generic timeout.

The way this timeout works is by inserting the timeout trigger when setting state to {?PEER_READ, ?SHUTDOWN} so this timeout will protect us from staleness in both cases of terminate, {ok, Checkpoint}} and {_, ok} in the MLD shutdown handler. Because both success/3 and shutdown_checkpoint/2 set next state to {?PEER_WRITE, ?SHUTDOWN} or {?PEER_WRITE, ?SHUTDOWN_CHECKPOINT} with the {write, IoData} event.

UPDATE:

• PR now only handles {?PEER_READ, ?SHUTDOWN}
• We use generic timeout instead of state timeout, so timeout always triggers regardless of state changes.

[CLAassistant]

All committers have signed the CLA.

[tak30]

I would be in favour of using a generic timeout just to ensure we really close the socket

[raz-adroll]

Also, can we log something if we receive the shutdown message?