docs(reference): import Dune: Awakening server-manager references

Phase 2 references for the host-agent Dune adapter, moved out of volatile /tmp
into docs/reference-repos/ (per Commander). Three upstream projects, .git +
node_modules + compiled binaries stripped (16MB source). Nested AI-instruction
files (.claude/, CLAUDE.md) removed so they don't pollute Corrosion sessions.

- icehunter/    dune-admin (Go+React) — 4 control planes; SETUP_DOCKER.md is the
                closest analog to our agent's Dune docker control plane (compose
                lifecycle, docker logs, RabbitMQ-via-exec, dune Postgres schema)
- adainrivers/  Rust/Tauri desktop — SSH+k8s BattleGroup control, maintenance
                daemon, in-game admin console (Rust idiom reference)
- the4rchangel/ Node web UI replacing battlegroup.bat — matches the Commander's
                Hyper-V self-host path + game-config schema

See docs/reference-repos/README.md for the full index + how we use each.

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

docs/reference-repos/adainrivers/release-notes/0.3.8.md

@@ -0,0 +1,14 @@
+Two reported issues against v0.3.7 — both fixed.
+
+## Reinstall step
+
+After updating the desktop app, click **Install / Update** on the Management Service card so the refreshed on-host service binary is pushed.
+
+## Fixed
+
+- **`Install management service` failed with `install: No such file or directory` when SSH'ing in as `root`** on Ubuntu 24.04+. Root cause: the install-init and write-token steps piped base64-decoded content into `sudo install /dev/stdin <dest>`, and sudoers' `Defaults use_pty` (default on noble) allocates a pty for the child process which severs the upstream pipe. The same flow worked when SSH'ing as `dune` because that account's sudoers entry bypasses `use_pty`. Both steps now stage their payload to a real temp file via `mktemp` first, then `sudo install <tmp> <dest>`, sidestepping the pty entirely.
+- **Director tunnel button grayed out even on a fully running BattleGroup**. We were parsing the vendor wrapper's `battlegroup status` text table for phase information, but newer Funcom operator builds changed the row vocabulary (the "Status" column now shows the partial world name, "Director" shows an `N/M` ready ratio, "Uptime" shows a `Healthy` / `Unhealthy` word instead of a duration). None of those values matched our phase allowlist, so the UI thought the BattleGroup was not started. We now read phase / serverGroupPhase / directorPhase / databasePhase / uptime / per-server stats directly from the BattleGroup CR's `status` object via `kubectl get battlegroup ... -o json`, which is the stable Kubernetes schema and immune to future wrapper text format changes.
+
+## Changed
+
+- Server status reads no longer go through the vendor wrapper. Start, stop, restart, and update lifecycle actions still use the wrapper (where it's the source of truth for orchestration).