Phase 7: netfox + godot-jolt stack upgrade

Stack installed:
- netfox v1.35.3 (core + extras + noray + internals)
- godot-jolt v0.16.0-stable

Architecture:
- Server: ENet transport (works headless, no netfox deps)
- Client/Editor: netfox rollback (RollbackSynchronizer, TickInterpolator)

New/modified:
- docs/migration-netfox-plan.md — migration architecture
- scripts/network/network_manager.gd — netfox-aware ENet fallback
- scripts/network/player.gd — clean base player
- client/characters/player_netfox.gd — rollback player w/ WeaponManager
- client/characters/input/player_net_input.gd — BaseNetInput subclass
- client/characters/character/fps_character_controller.gd — netfox input feed
- client/weapons/ — weapon data, registry, TacticalWeaponHitscan, WeaponManager
- client/scripts/round_replicator.gd — client-side round state bridge
- server/scripts/round_manager.gd — improved state machine
- server/scripts/plugin_api/plugin_manager.gd — refined plugin system
- config: enemy_tag, ally_tag for meatball targeting

Removed: old C++ SimulationServer GDExtension (replaced by netfox rollback)

server/server-browser-api/venv/lib/python3.11/site-packages/fastapi/concurrency.py

@@ -0,0 +1,41 @@
+from collections.abc import AsyncGenerator
+from contextlib import AbstractContextManager
+from contextlib import asynccontextmanager as asynccontextmanager
+from typing import TypeVar
+
+import anyio.to_thread
+from anyio import CapacityLimiter
+from starlette.concurrency import iterate_in_threadpool as iterate_in_threadpool  # noqa
+from starlette.concurrency import run_in_threadpool as run_in_threadpool  # noqa
+from starlette.concurrency import (  # noqa
+    run_until_first_complete as run_until_first_complete,
+)
+
+_T = TypeVar("_T")
+
+
+@asynccontextmanager
+async def contextmanager_in_threadpool(
+    cm: AbstractContextManager[_T],
+) -> AsyncGenerator[_T, None]:
+    # blocking __exit__ from running waiting on a free thread
+    # can create race conditions/deadlocks if the context manager itself
+    # has its own internal pool (e.g. a database connection pool)
+    # to avoid this we let __exit__ run without a capacity limit
+    # since we're creating a new limiter for each call, any non-zero limit
+    # works (1 is arbitrary)
+    exit_limiter = CapacityLimiter(1)
+    try:
+        yield await run_in_threadpool(cm.__enter__)
+    except Exception as e:
+        ok = bool(
+            await anyio.to_thread.run_sync(
+                cm.__exit__, type(e), e, e.__traceback__, limiter=exit_limiter
+            )
+        )
+        if not ok:
+            raise e
+    else:
+        await anyio.to_thread.run_sync(
+            cm.__exit__, None, None, None, limiter=exit_limiter
+        )