Client-side position sync + prediction/reconciliation

Real client-side networking for the player controller, verified with a
genuine two-process integration test (an actual dedicated server process
+ a real ENet-connected client), per docs/architecture/networking.md §5.

scripts/network_client.gd: NetworkClient -- predicts movement locally on
input (same movement constants as scripts/player_spawner.gd), sends input
to the server via RPC, and reconciles against the server's authoritative
snapshots when they diverge beyond a threshold.

scripts/network_receiver_proxy.gd + network_player_proxy.gd: the RPC
plumbing that makes this actually route correctly, which took real
debugging to get right:
- Godot's rpc_id() resolves the receiving node by matching the sending
  node's NodePath between peers. scripts/server.gd is the SceneTree root
  node "Server" when the dedicated server runs, and spawns each player as
  "Player_<peer_id>" under itself -- so the client needs its own node
  tree at those same paths purely for RPC routing.
- Godot also enforces an RPC "node checksum": both peers must declare the
  *same set* of @rpc methods at a matching path, or every RPC at that
  path fails, not just a genuinely missing one. Every @rpc method on
  scripts/server.gd's root and scripts/player_spawner.gd had to be
  mirrored here (mostly as no-op stubs), not just the ones this class
  actually calls.

Known, explicitly-scoped simplification (stated in the code, not hidden):
scripts/server.gd's snapshot doesn't currently echo back a per-peer "last
input seq processed", so reconciliation corrects against the most recent
prediction rather than true tick-exact rollback-and-replay
(networking.md §5.2 step 4). A real follow-up once the server echoes
that ack, not faked here.

Verified: real two-process test (godot --headless --path . scenes/
server.tscn as an actual persistent process, a second process running
scripts/smoke_test_network_client.gd against it over real ENet) --
4/4 passing: connects, predicts locally before any server round trip,
sends real input, and a deliberately-desynced position (moved 9999 units
away) gets corrected by the very next real snapshot. Full existing
client suite (10 suites) + 61 backend pytest still passing.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01M6rUEynmv2QmweocAQYARN

docs/testing/test_inventory.md

@@ -53,7 +53,8 @@ these yet):
 
 ## Godot client — `scripts/smoke_test_*.gd`
 
-**10 smoke tests merged into `main`.** Each needs a live backend —
+**11 smoke tests merged into `main`.** Ten need a live backend, one
+(`smoke_test_network_client.gd`) needs a live dedicated server instead —
 `AGP_API_BASE_URL` env var, falling back to the
 `application/agp/api_base_url` project setting (default
 `http://127.0.0.1:8000`) if unset.
@@ -70,6 +71,7 @@ these yet):
 | `smoke_test_test_level.gd` | Loads `scenes/worlds/test_level.tscn` directly (the level is player-agnostic and self-contained per `docs/architecture/pc2_test_level.md` §4) and confirms it has real content: 5 props, a `DirectionalLight3D` ("Sun"), and a `WorldEnvironment` — not an empty void. |
 | `smoke_test_player.gd` | Loads `scenes/game.tscn` and confirms the player exists as a real `CharacterBody3D` with a `CollisionShape3D`+shape, a `CapsuleMesh`, and spawns at the level's `PlayerSpawn` marker position. Also asserts `JUMP_VELOCITY` is positive and a `Camera3D` exists and is `current` — both were real bugs found 2026-08-20/21 (jump velocity was negative, pushing the player down instead of up; no camera existed at all, so the game was unplayable even though the character could technically move). Structural checks only — real movement/jump *direction* correctness was verified separately via a throwaway headless physics simulation during development, not by an automated test that's part of this suite. |
 | `smoke_test_enter_world.gd` | The full PC2 path end to end: register → seed a real world → load `world_list.tscn` → click the real "Enter World" button on a real card (found by walking the card's children, not by index, so it doesn't silently pass if the card layout changes) → confirms `scenes/game.tscn` actually becomes `current_scene` with the player spawned at the level's `PlayerSpawn` marker → sends a real `KEY_ESCAPE` input event (not a direct method call) and confirms it returns to `world_list.tscn`. This is the test that proves "browse → enter → walk around → leave" actually works as one continuous flow, not just that each piece works in isolation. |
+| `smoke_test_network_client.gd` | **Real two-process integration test** for `scripts/network_client.gd` (client-side prediction + reconciliation, docs/architecture/networking.md §5) — needs an actual dedicated server running first: `godot --headless --path . scenes/server.tscn &`, *not* the usual HTTP backend. Connects a real `NetworkClient` over ENet, confirms local input prediction moves the player immediately (no waiting on a round trip), confirms real snapshots arrive and get processed, then deliberately desyncs the predicted position by 9999 units and confirms the very next snapshot corrects it back — the actual proof reconciliation logic works, not just that it never had to fire. Known simplification, stated in the script's own docstring: the server's snapshot doesn't currently echo back a per-peer "last input seq processed," so reconciliation corrects against the *most recent* prediction rather than true tick-exact rollback-and-replay (docs/architecture/networking.md §5.2 step 4) — a real, scoped follow-up once the server echoes that ack, not something faked here. |
 
 Run (once, to build the class cache): `godot --headless --path . --import`,
 then for each smoke test: `godot --headless --path . --script