Ensure instance monitoring objects (alive/connected) exist

[krobipd]

Symptom

The devices instance is shown red in the admin Instances tab for a few seconds every time the tab is opened (or after a browser refresh), then turns green. It happens only for devices, not for other adapters, and nothing is logged — even in debug. Reported in #603 and in the forum (topic 84587); the workaround users found was to delete and re-create the instance.

Root cause

On the affected installations the instance monitoring objects system.adapter.devices.<n>.alive and system.adapter.devices.<n>.connected are missing, while their state values still exist (js-controller keeps writing them via the heartbeat).

When the Instances tab loads, its object-based state snapshot (getForeignStates('system.adapter.*')) skips the object-less states, so alive/connected are briefly absent from the admin's cache and getInstanceStatus() evaluates !alive || !connected → red. The live state subscription then delivers a heartbeat and the row turns green.

This is specific to devices: other connectionType: none daemon adapters (e.g. javascript) still have both objects. The current code neither deletes nor recreates them, so once they are gone (older version / upgrade) the situation persists.

Fix

Recreate the two objects idempotently on start with setForeignObjectNotExistsAsync, modeled on the standard instance-state objects (boolean, role indicator.state). Healthy instances are left untouched; affected instances self-heal on the next start — no data loss, unlike delete-and-recreate.

How it was verified

On a live system with the objects missing (red confirmed in Safari and Firefox):

• Creating the two objects → the instance turns green immediately and stays green across tab switches and force-refresh.
• The objects survive an adapter restart — nothing removes them again.
• getForeignStates('system.adapter.devices.0.*') returned the states again once the objects existed (it was empty before).

npm run build:backend and eslint pass.

Addresses #603.