feat(reservoir): warm-start grid with analytical lead values

Pre-populate all LeadGrid cells with arcsin(vPerp/bulletSpeed) so the
grid starts warm instead of cold, avoiding the early-round 0-data
fallback to -999 (no-lead) firing.

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

SNNBot_garage/src/reservoir.nim

@@ -16,8 +16,18 @@ type
   LeadGrid* = object
     cells: array[VPERP_BINS * DIST_BINS, GridCell]  # 17 × 8 = 136 cells
 
+const BULLET_SPEED = 14.0  # power 2 bullet speed in pixels/tick
+
 proc initLeadGrid*(): LeadGrid =
-  result = LeadGrid()
+  for vBin in 0..<VPERP_BINS:
+    for dBin in 0..<DIST_BINS:
+      let vPerp = float(vBin) - 8.0
+      let sinVal = clamp(vPerp / BULLET_SPEED, -1.0, 1.0)
+      let offsetRad = arcsin(sinVal)
+      let idx = vBin * DIST_BINS + dBin
+      result.cells[idx].sumSin = sin(offsetRad)
+      result.cells[idx].sumCos = cos(offsetRad)
+      result.cells[idx].count = 1
 
 proc cellIndex(vPerp: float, distance: float): int {.inline.} =
   let vBin = clamp(int(vPerp + 8.5), 0, VPERP_BINS - 1)