fix: Refresh SIP settings on server config change

[edospadoni]

Summary

NethLink parsed config.production.js only at credential login and cached sipHost and sipPort in the account. When NethVoice reassigned the SIP port, typically after an update, those cached values went stale and the PhoneIsland could no longer register with Janus. The user was left with a permanent connection error banner and the only workaround was to log out and log back in. Restarting the app was not enough, because autologin reused the saved account as-is whenever the JWT was still valid.

Two changes fix this:

• At startup, autologin re-reads the server config in both the valid-token and the expired-token path, and persists the new values.
• While running, two independent triggers detect a change and reload the PhoneIsland so it re-registers with the new port. The first is the socket reconnection that follows a server restart. The second is the webrtc_down alert, raised on every failed registration, which acts as a safety net when the config endpoint comes back after the socket. Config checks from the second trigger are debounced to 10 seconds.

The reload is deferred to the end of an active call, so a call in progress is never dropped. A failed config fetch is swallowed and the cached values are kept, so a server that is still restarting can never break autologin.

Also removes PhoneIslandController.reconnect(), its RECONNECT_PHONE_ISLAND event and the matching renderer listener, which had no callers.

How to test

Run the app with DEV=true npm run dev. Main process logs go to stdout.

On the NS8 node, replacing nethvoice1 and 20027 with the instance values:

mid=nethvoice1

# A - advertise a port Asterisk is not listening on
runagent -m $mid podman exec nethcti-ui \
  sed -i "s/20027/20099/" /app/public/config/config.production.js

# B - restart the CTI server, as a module update does
runagent -m $mid systemctl --user restart nethcti-server

# C - restore, and undo any of the tests below
runagent -m $mid systemctl --user restart nethcti-ui

A touches only the container writable layer. Do not edit ASTERISK_SIP_PORT instead: the SIP proxy dispatcher targets it.

#	Scenario	Steps	Expected
1	Port changes while the app is closed	quit, A, start	server SIP config changed at startup, then the error banner
2	Port changes while the app runs	A, B	detection, reloading PhoneIsland, banner; C clears it in ~10s with no clicks
3	Plain reconnection, no regression	B alone	no detection, no reload, stays registered
4	Port changes during a call	call, then A, B	call survives with audio, reload deferred to call end, one reload on hangup

All four verified against a live NethVoice instance.

[github-actions]

Automatic builds from https://github.com/NethServer/nethlink/actions/runs/35213489796.
Commit: 9fc94ac

Name	Platform	Link
win-app.exe	Windows (x64)	Link
macos-app-x64.dmg	MacOS (x64)	Link
macos-app-arm64.dmg	MacOS (arm64)	Link
linux-app.AppImage	Linux (x64)	Link

[andrea-marchionni]

Verified.