Add DevControlBot garage: skeleton bot with release/debug build script

A minimal Tank Royale bot that boots and stands still. Body, gun and radar
are set to white once at init (team convention: white = not programmed yet);
the tick loop only calls go().

- DevControlBot/ holds the single bot module, its metadata JSON, the
  BotLauncher entry script and the nimble file
- DevControlBot.sh is a thin wrapper around the runBot nimble task:
  release build by default, --debug for the classic diagnostic build
- dependencies resolve at run time from the global nimble store via
  `nimble path`; optional .env (gitignored) supplies env vars such as
  SERVER_URL, with existing environment variables taking precedence
- build happens in a throwaway mktemp dir removed on exit; no artifacts

DevControlBot_garage/DevControlBot/DevControlBot.sh

@@ -0,0 +1,92 @@
+#!/usr/bin/env bash
+# BotLauncher entry point for DevControlBot.
+#
+# Thin wrapper: all build/run/cleanup logic lives in the `runBot` nimble task
+# of DevControlBot.nimble. This script only locates itself, moves next to the
+# .nimble so Nimble finds it, forwards caller arguments, calls the task and
+# propagates the real exit code. Signals are forwarded to Nimble so the task's
+# own EXIT/INT/TERM cleanup always runs.
+#
+# Usage: DevControlBot.sh [--debug] [<metadata.json>] [extra args for the bot]
+#
+#   (no args)   RELEASE build + run with the bundled DevControlBot.json
+#   (no flag)   RELEASE build (nim c -d:release), quiet output
+#   --debug     classic DEBUG build (plain nim c): compiler hints + diagnostics
+#
+# --debug is a build-mode flag for this wrapper only: it is stripped here and
+# never reaches the bot binary or nimble's task arguments. It travels to the
+# runBot task as $DEVCONTROLBOT_DEBUG (exported), because an env var cannot be
+# confused with a bot argument, cannot collide with the status file mechanism
+# and keeps the task's argument parsing untouched. Repeated --debug is harmless.
+#
+# Configuration (SERVER_URL, SERVER_SECRET, ...) is NOT handled here at all: the
+# API's start() reads those from the environment, and the optional .env file is
+# loaded by the runBot task, so it also works with a direct `nimble runBot`.
+set -uo pipefail
+set -m   # job control: the background job gets its own process group
+
+SCRIPT_DIR="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" && pwd)"
+cd "$SCRIPT_DIR"
+
+usage() {
+  cat <<EOF
+Usage: $(basename "$0") [--debug] [<metadata.json>] [extra args for the bot]
+
+  --debug    classic debug build: shows compiler hints and diagnostics
+              (default is a quiet RELEASE build, nim c -d:release)
+  -h,--help   show this help
+
+Examples:
+  $(basename "$0")                       # release build + run, bundled DevControlBot.json
+  $(basename "$0") my.json               # release build + run with custom metadata
+  $(basename "$0") -h                    # this help
+  $(basename "$0") --debug               # debug build
+  $(basename "$0") --debug my.json       # debug build with custom metadata
+EOF
+}
+
+# Split caller args: everything except --debug is forwarded verbatim.
+BUILD_MODE=release
+FORWARDED=()
+for a in "$@"; do
+  case "$a" in
+    --debug) BUILD_MODE=debug ;;
+    *) FORWARDED+=("$a") ;;
+  esac
+done
+set -- ${FORWARDED+"${FORWARDED[@]}"}
+
+# No arguments is the normal case: build and run with the bundled metadata.
+# Usage is printed ONLY for an explicit -h/--help.
+case "${1:-}" in
+  -h|--help) usage; exit 0 ;;
+esac
+
+export DEVCONTROLBOT_DEBUG="$BUILD_MODE"
+
+# The runBot task's shell always exits 0 (otherwise Nimble raises a NimScript
+# exception with a stack trace and a dump of the whole script) and writes the
+# bot's real exit code here instead. We read it back and exit with it, so
+# BotLauncher still sees the faithful status: 1 = no game server on 7654,
+# 2 = a dependency is missing, 3 = compile failure, anything else = the bot's
+# own code, 0 = clean run.
+STATUS_FILE="$(mktemp "${TMPDIR:-/tmp}/devcontrolbot.status.XXXXXX")"
+export DEVCONTROLBOT_STATUS_FILE="$STATUS_FILE"
+trap 'rm -f "$STATUS_FILE"' EXIT
+
+# Forward to the whole job process group (nimble + the shell it spawned) so the
+# task's own EXIT/INT/TERM cleanup runs. No cleanup logic here.
+forward() { kill -s "$1" -- "-$NIMBLE_PID" 2>/dev/null; }
+trap 'forward TERM' TERM
+trap 'forward INT' INT
+
+nimble runBot ${1+"$@"} &
+NIMBLE_PID=$!
+wait "$NIMBLE_PID"
+rc=$?
+trap - TERM INT
+# Prefer the code the run actually produced; fall back to Nimble's own status if
+# the task died before it could report (e.g. Nimble itself failed).
+real_rc="$(head -n 1 "$STATUS_FILE" 2>/dev/null | tr -d '[:space:]')"
+[ -n "$real_rc" ] || real_rc=$rc
+exit "$real_rc"