scaleway/compute-instance v2.3.0
Adds a new optional cockpit input to scaleway/compute-instance (ADR-0102): when set, the module writes /etc/alloy/config.alloy and installs Grafana Alloy on first boot, scraping pigeon-cli's local Prometheus metrics endpoint (noisypigeon/pigeon-cli ADR-0092) and tailing its JSONL log (pinned to a stable /var/log/pigeon/pigeon.jsonl path via a new PIGEON_LOG_DIR export), forwarding both to a Scaleway Cockpit project's metrics/logs sources. A whole-instance node_exporter scrape is included as a low-cost addition once Alloy is installed anyway.
cockpit is null by default (fully opt-in) and only used when profile = "pigeon-cli", matching the existing buckets/keyring_entries gating pattern. Applying this to an existing instance will force-replace it (cloud-init content change, per this module's existing replace_triggered_by behavior) — something to call out explicitly in whichever leaf PR actually sets cockpit.
The actual scaleway_cockpit_source/scaleway_cockpit_token resources and wiring cockpit = {...} into a real leaf are a deliberate follow-up, not part of this PR — they need this module's new tag to reference.
See docs/adr/0102-compute-instance-cockpit-alloy.md for full rationale, including one unverified assumption (the Loki-side auth header) flagged explicitly in the ADR's Out of scope section.