feat(PPO_Bot): command abstraction layer — goto/aimTo controllers (#24)

- Add gotoTick/aimToTick controller functions (#25)
- Update network dims: actor 5→6, state 42→44 (#26)
- Rewrite mapActions for 6-dim command space (#27)
- Delete stale weight files (shape mismatch)
- Fix existing tests for new signatures

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

PPO_Bot/state_vector.nim

@@ -1,4 +1,4 @@
-## State vector builder — produces 42-float normalized tensor for PPO policy.
+## State vector builder — produces 44-float normalized tensor for PPO policy.
 ## No bot API imports; takes plain BotState + EnemyTracker structs.
 
 import std/math
@@ -16,10 +16,12 @@ type
     gunHeat*: float64
     arenaWidth*, arenaHeight*: float64
 
-proc buildStateVector*(bot: BotStateData; enemy: EnemyTracker): Tensor[float32] =
-  ## Build the 42-float normalized state tensor.
+proc buildStateVector*(bot: BotStateData; enemy: EnemyTracker;
+                       remainingGotoDistance: float64 = 0.0;
+                       remainingGunAngle: float64 = 0.0): Tensor[float32] =
+  ## Build the 44-float normalized state tensor.
   ## All values clipped to roughly [-1, 1] via division by physical maxima.
-  result = zeros[float32](42)
+  result = zeros[float32](44)
 
   let aW = bot.arenaWidth
   let aH = bot.arenaHeight
@@ -85,3 +87,7 @@ proc buildStateVector*(bot: BotStateData; enemy: EnemyTracker): Tensor[float32]
       result[base + 2] = float32(enemy.history[i].direction / 360.0)
       result[base + 3] = float32(enemy.history[i].speed / 8.0)
     # else: remain 0.0 (pad)
+
+  # --- Goto controller inputs (indices 42-43) ---
+  result[42] = float32(remainingGotoDistance / diag)
+  result[43] = float32(remainingGunAngle / 180.0)