feat: VectorField node — grid-based motion vector visualization from tracker output

[Copilot]

Adds a new VectorField node under the Visual menu that consumes MultiObjectTracking JSON output and renders a per-cell motion vector field overlaid on the video frame.

New node: node/VisualNode/node_vector_field.py

Inputs: IMAGE + JSON (tracker: track_ids, bboxes)
Output: IMAGE with arrow overlay

Vector field rendering

Each grid cell accumulates velocity vectors from all tracked objects whose trajectory midpoints fell within it. Per-cell arrow:

• Direction → average velocity vector
• Length → proportional to relative speed (0 → max observed)
• Thickness → 1–6 px, proportional to relative speed
• Color → blue (slow) → green → red (fast) via HSV mapping

UI controls

Control	Range	Purpose
Cell size (px)	10–200	Square grid resolution; field recalculates on change
Period mode	Minutes / Hours / Infinite	Round-robin retention unit
Duration	1–60 (min) or 1–24 (h)	History window depth
Blend Alpha	0.0–1.0	Overlay weight vs. background frame

Round-robin history

Position history is stored per track_id as a deque of (cx, cy, timestamp) tuples. Old entries are pruned on every update based on the configured retention window. Velocity is computed in display-pixel/second between consecutive frames, scaled to display resolution.

Style registration: node_editor/style.py

VectorField added to VIZ list → appears in the Visual menu with the existing pink-pastel theme.