Keep the disconnect record right when a display change lands after its timeout

[didriksg]

What & why

setEnabled stops waiting for CGCompleteDisplayConfiguration after 10 s, but the commit can still land later (#33 saw 29.5 s). Three things went wrong when it did. A disconnect that timed out stored no record, so a late landing left the display dark with no Reconnect row and no blackout rescue. A reconnect that timed out kept its record, so a late landing met that record in reconcile and was switched straight back off. And the DDC hold ended when the wrapper gave up, so DDC traffic could start again under a commit that was still in WindowServer.

Both timeouts now keep the record and watch the display for 30 s, reconcile leaves it alone meanwhile, and enumeration decides. The hold now ends when the transaction returns. After a late reconnect the input is not written back. Stacked on #218, which changes the same functions. Found while reading the backend in #209 and #210.

How tested

• make check.
• A dev build with the wrapper set to 0.01 s, so every commit "times out" and lands about 100 ms later, on a MacBook with one external: the disconnect kept its record and logged landed late, and the reconnect logged landed late, record dropped and stayed connected, with no re-apply from reconcile.
• The DDC hold change has nothing to see from outside. It only changes when the hold ends.

Checklist

• Builds locally
• Added/updated Crisp/Resources/Localizable.xcstrings for any new user-facing strings (none)
• Screenshot or short clip for any UI change (none)