OTA only needs the local node to send the startOTA admin request and then
streams firmware directly to the loader over WiFi; it never reads the node DB.
On a device with a large node DB the initial dump can stall and the firmware
closes the API connection before it processes the OTA request, so OTA silently
fails. Imply --no-nodes for --ota-update / --reboot-ota.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

…n in commit, and allow specifying a version, as well as allowing dispatch from other repositories.

…otobufs

Improve update_protobufs workflow

…ip-node-db

Skip node DB fetch during --ota-update / --reboot-ota

The container build now logs in and pushes only on push events, so a merge queue run builds the image without publishing it.

Run CI and the container build on merge_group events