feat(server): task 6.1/6.3 — dedicated server export preset and a real CLI surface

6.1: "Linux Dedicated Server" preset (dedicated_server=true,
custom_features="dedicated_server") mirroring the existing training
preset, plus run/main_scene.dedicated_server so the server binary reaches
its own entry point with no flag. Builds: an 85MB Linux x86_64 binary,
gitignored like the training one.

6.3: scripts/server_config.gd declares every server flag once - name,
type, default, section, help - and one parser turns that into parsing,
type checking, range validation, config-file backing and --help. The
flags had grown to ~30 across server_boot.gd and networked_match.gd, each
parsed inline with begins_with, none documented, and an unrecognised flag
was SILENTLY IGNORED: --max-clientss=8 ran a server on the default cap
and said nothing. Unknown flags, missing values, wrong types, duplicates
and out-of-range values are now hard errors, reported all at once.

Precedence is command line > config file > default. server_boot.gd parses
strictly because it owns the whole command line; networked_match.gd reads
the same declaration leniently because it is one consumer of an argv the
smoke harnesses also fill with --role= and --drive-seconds=. Nothing is
lost - every server flag is declared, so the strict pass already caught
any typo before the match scene re-reads its own.

13 unit tests covering the precedence order, the typo rejection that
motivated this, --no-<bool> not double-listing in --help, and --help
documenting every flag asserted against the declaration rather than a
hand-kept list. Verified end to end: --help prints, a typo'd flag refuses
to start, and the plain/replay-log/late-joiner smoke scenarios still pass.

Game/scripts/server_boot.gd

@@ -17,6 +17,7 @@ const LOG_LEVELS := {"debug": 0, "info": 1, "warn": 2, "error": 3}
 var _boot_ms := 0
 var _last_physics_frame := 0
 var _log_level := 1   # info
+var config: ServerConfig = null
 var _watchdog_armed := false   # skip the first _process(): engine startup scheduling can batch several physics frames before the first idle frame runs, which isn't a real overrun
 
 
@@ -24,21 +25,28 @@ func _ready() -> void:
 	_boot_ms = Time.get_ticks_msec()
 	Engine.max_fps = 60   # a server never renders; this just caps the idle-frame poll rate so it doesn't spin
 
-	var port := NetworkManager.DEFAULT_PORT
-	var max_clients := NetworkManager.MAX_CLIENTS
-	for arg in OS.get_cmdline_user_args():
-		if arg.begins_with("--port="):
-			port = int(arg.substr("--port=".length()))
-		elif arg.begins_with("--max-clients="):
-			max_clients = int(arg.substr("--max-clients=".length()))
-		elif arg.begins_with("--log-level="):
-			var level_name := arg.substr("--log-level=".length())
-			if LOG_LEVELS.has(level_name):
-				_log_level = LOG_LEVELS[level_name]
-			else:
-				_log("error", "bad_log_level", {"given": level_name, "valid": LOG_LEVELS.keys()})
-				get_tree().quit(1)
-				return
+	# Task 6.3. This process owns the whole command line, so it parses STRICTLY:
+	# an unknown flag or an out-of-range value stops the server with a message
+	# rather than starting one that silently ignores half of what it was told.
+	config = ServerConfig.parse(OS.get_cmdline_user_args())
+	if config.help_requested:
+		print(ServerConfig.help_text())
+		get_tree().quit(0)
+		return
+	if not config.is_valid():
+		# Straight to stderr-ish plain print rather than through _log: the log
+		# level itself may be one of the things that failed to parse, and an
+		# operator running this by hand needs to see every problem at once, not
+		# the first one.
+		printerr("cosmic-clash-server: refusing to start")
+		for problem in config.errors:
+			printerr("  %s" % problem)
+		printerr("try --help")
+		get_tree().quit(1)
+		return
+	var port := int(config.get_value("port"))
+	var max_clients := int(config.get_value("max-clients"))
+	_log_level = LOG_LEVELS[String(config.get_value("log-level"))]
 
 	NetworkManager.client_connected.connect(_on_client_connected)
 	NetworkManager.client_disconnected.connect(_on_client_disconnected)