feat(p1x): de-hardcode VLM models/endpoints to vlm_config (DGX-ready)

Migre les call-sites VLM serveur vers la configuration centrale pour
fonctionner sur DGX (tunnel Ollama 11434), où gemma4:* est absent et le
port Docker 11435 est mort.

- task_planner, replay_verifier, domain_context, ir_builder, resolve_engine
  (popup): modele -> vlm_config.get_vlm_model(), defaut 11435 -> 11434
  (override GEMMA4_PORT legacy conserve)
- resolve_engine (grounding bbox x2): nouvel helper
  vlm_config.get_bbox_grounding_model() (var dediee RPA_BBOX_GROUNDING_MODEL,
  fallback RPA_GROUNDING_MODEL puis qwen2.5vl:7b-rpa) -> desambiguise le
  conflit D5-v3b, bbox_2d + num_ctx 4096 preserves
- safety_checks_provider: defaut -> get_vlm_model(), override
  RPA_SAFETY_CHECKS_LLM_MODEL preserve
- ui_detector: default_factory + resolution lazy (corrige aussi un gel a
  l'import), pas d'appel reseau a l'import
- field_extractor: property lazy via vlm_config

TDD strict (RED->GREEN), 305 tests verts, tests mockes HTTP (zero dependance
DGX reel), aucun alias Ollama.

Hors perimetre (arbitrage Dom): client Lea agent_v1/executor.py (gele),
chemin V4 observe_reason_act (RPA_REASONING_MODEL), core/config.py defaults.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

tests/unit/test_vlm_grounding_profile.py

@@ -96,6 +96,41 @@ def test_grounding_profile_gemma4_triggers_think_false(monkeypatch):
     assert p["think"] is False  # gemma4 needs think=false
 
 
+# ────────────────────────────────────────────────────────────────────────────
+# get_bbox_grounding_model (D5-v3b : modèle bbox_2d dédié, désambiguïsé)
+# ────────────────────────────────────────────────────────────────────────────
+
+
+@pytest.mark.unit
+def test_bbox_grounding_model_default(monkeypatch):
+    """Sans env, défaut = DEFAULT_GROUNDING_FALLBACK (qwen2.5vl:7b-rpa, présent DGX)."""
+    monkeypatch.delenv("RPA_BBOX_GROUNDING_MODEL", raising=False)
+    monkeypatch.delenv("RPA_GROUNDING_MODEL", raising=False)
+    from core.detection.vlm_config import get_bbox_grounding_model, DEFAULT_GROUNDING_FALLBACK
+
+    assert get_bbox_grounding_model() == DEFAULT_GROUNDING_FALLBACK == "qwen2.5vl:7b-rpa"
+
+
+@pytest.mark.unit
+def test_bbox_grounding_model_dedicated_env(monkeypatch):
+    """RPA_BBOX_GROUNDING_MODEL est prioritaire."""
+    monkeypatch.setenv("RPA_BBOX_GROUNDING_MODEL", "qwen2.5vl:32b")
+    monkeypatch.setenv("RPA_GROUNDING_MODEL", "qwen3.5:9b")
+    from core.detection.vlm_config import get_bbox_grounding_model
+
+    assert get_bbox_grounding_model() == "qwen2.5vl:32b"
+
+
+@pytest.mark.unit
+def test_bbox_grounding_model_legacy_compat(monkeypatch):
+    """Sans la var dédiée, RPA_GROUNDING_MODEL reste honoré (rétrocompat)."""
+    monkeypatch.delenv("RPA_BBOX_GROUNDING_MODEL", raising=False)
+    monkeypatch.setenv("RPA_GROUNDING_MODEL", "qwen2.5vl:7b-rpa")
+    from core.detection.vlm_config import get_bbox_grounding_model
+
+    assert get_bbox_grounding_model() == "qwen2.5vl:7b-rpa"
+
+
 # ────────────────────────────────────────────────────────────────────────────
 # _extract_first_json_object
 # ────────────────────────────────────────────────────────────────────────────