perf(linux): build configs in parallel (default)

Linux built the 3 configs sequentially (~3x a single config); Windows
already parallelizes. Run them concurrently via ThreadPoolExecutor,
splitting make -j across configs to avoid CPU oversubscription.
Symlink setup serialized with a lock. --no-parallel restores
sequential. Brings Linux wall time close to Windows.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

tools/linux/build_plugin_linux.py

@@ -15,8 +15,14 @@ import os
 import shutil
 import subprocess
 import sys
+import threading
+from concurrent.futures import ThreadPoolExecutor
 from pathlib import Path
 
+# Symlink creation is shared across configs; serialize it so parallel
+# per-config builds don't race recreating the bundle symlinks.
+_SYMLINK_LOCK = threading.Lock()
+
 ROOT = Path(__file__).resolve().parents[2]
 
 OVERLAY = ROOT / "tools" / "linux" / "overlay"
@@ -158,7 +164,9 @@ def _build_one(
     bundle_dir  = layout["bundle_dir"]
 
     # Symlinks are not committed to git; recreate them on Linux if missing.
+    # Serialized: concurrent per-config builds would race here.
     if zip_version in VERSION_LAYOUT:
+        with _SYMLINK_LOCK:
             _ensure_bundle_symlinks(zip_version, bundle_dir, vendor_7zip)
 
     cfg = CONFIGS[cfg_name]
@@ -313,6 +321,18 @@ def main() -> int:
         help="Copy built DLLs to <repo>/dist/<config> instead of "
              "<repo>/plugins/<config>",
     )
+    parser.add_argument(
+        "--parallel",
+        action="store_true",
+        default=True,
+        help="Build configs concurrently (default: True)",
+    )
+    parser.add_argument(
+        "--no-parallel",
+        action="store_false",
+        dest="parallel",
+        help="Build configs one at a time",
+    )
     args = parser.parse_args()
 
     _require_tool("make")
@@ -329,6 +349,35 @@ def main() -> int:
 
     wanted = list(CONFIGS.keys()) if "all" in args.configs else args.configs
 
+    if args.parallel and len(wanted) > 1:
+        # Split make jobs across configs so 3 concurrent makes don't
+        # oversubscribe the CPU (mirrors the Windows parallel build).
+        per_jobs = max(1, jobs // len(wanted))
+        print(
+            f"[linux] Building {len(wanted)} configs in parallel "
+            f"(make -j{per_jobs} each)"
+        )
+        codes: dict[str, int] = {}
+        with ThreadPoolExecutor(max_workers=len(wanted)) as ex:
+            futs = {
+                ex.submit(
+                    _build_one,
+                    args.zip_version,
+                    c,
+                    args.verbose,
+                    per_jobs,
+                    args.clean,
+                    args.cleanup_artifacts,
+                    args.dist,
+                ): c
+                for c in wanted
+            }
+            for fut, name in futs.items():
+                codes[name] = fut.result()
+        for name in wanted:
+            if codes[name] != 0:
+                return codes[name]
+    else:
         for cfg_name in wanted:
             code = _build_one(
                 args.zip_version,