Cleanup repo + harden offline install and watchdog

Repo hygiene:
- Remove dead scripts (.wait-for-display, .display-keepalive, .keep-screen-alive)
- Remove stale backup/scratch (install.sh.bak, New.txt, stale .pyc)
- Stop tracking machine-specific/generated state (config/app_config.json,
  playlists/server_playlist.json, src/player_auth.json, generated launchers)
- Prune duplicate wheel versions (one per package)
- Reorganize working_files -> dev/, dev/tests/, documentation/dev-notes/, scripts/
- Extend .gitignore; fix no-op rules in repo/.gitignore

Offline install (install.sh):
- Non-fatal system-dep step; per-package apt retry; .deb bundle install
- Skip pip self-upgrade offline; guard online fallbacks behind connectivity probe
- Dependency-free fetch_url (curl/wget/urllib); --no-reboot flag
- Fix corrupted literal backslash-bang escapes (shebangs, test -f, banner)
- Idempotent cron install (no more stacked @reboot lines)

Autostart / watchdog:
- start.sh: single-instance flock lock (.player_watchdog.lock) acquired only
  after a display is confirmed, preventing parallel watchdogs
- Set XDG_RUNTIME_DIR when missing; exit early if headless
- run_player.sh: activate .venv; stop_player.sh: clear lock file

start.sh

@@ -59,6 +59,13 @@ load_user_environment() {
     # Fallback: detect display manually
     log_message "Falling back to manual display detection..."
     
+    # Ensure XDG_RUNTIME_DIR is set. Wayland clients (and Kivy) need it, and a
+    # cron-launched instance may not have it. Default to the standard location.
+    if [ -z "$XDG_RUNTIME_DIR" ]; then
+        export XDG_RUNTIME_DIR="/run/user/$(id -u)"
+        log_message "Set XDG_RUNTIME_DIR=$XDG_RUNTIME_DIR"
+    fi
+    
     if [ -z "$DISPLAY" ] && [ -z "$WAYLAND_DISPLAY" ]; then
         # Try to detect Wayland display
         if [ -S "/run/user/$(id -u)/wayland-0" ]; then
@@ -96,8 +103,27 @@ load_user_environment() {
     return 0
 }
 
-# Load the user environment early
-load_user_environment
+# Load the user environment early. If there is no usable display, exit WITHOUT
+# taking the lock — otherwise a headless instance (e.g. cron before login) would
+# block the real desktop-autostart instance from ever running.
+if ! load_user_environment; then
+    log_message "No display server available — exiting so another instance can start"
+    exit 0
+fi
+
+# ============================================================
+# Single-instance guard (prevents parallel watchdogs)
+# ============================================================
+# Only ONE watchdog may run at a time. If XDG autostart, the cron fallback, and
+# a manual start all fire, the first one to reach this point wins; every other
+# instance exits immediately instead of spawning a second player.
+LOCK_FILE="$SCRIPT_DIR/.player_watchdog.lock"
+exec 200>"$LOCK_FILE"
+if ! flock -n 200; then
+    log_message "Another watchdog is already running (lock: $LOCK_FILE) — exiting"
+    exit 0
+fi
+log_message "Single-instance lock acquired: $LOCK_FILE"
 
 # Function to configure display output to Full HD (1920x1080)
 configure_display_resolution() {