feat: add database query and prediction diagnostic script

Add scripts/query_db.py with utilities to inspect the database:
- employee count and churn rate stats
- preview of first N employees
- on-demand ML prediction for a given employee id
- logged prediction stats (count, latest entry)

scripts/query_db.py

@@ -0,0 +1,80 @@
+import sys
+import os
+
+sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+from dotenv import load_dotenv
+load_dotenv()
+
+import pandas as pd
+from app.db.session import SessionLocal
+from app.db.models import Employee, Prediction
+from app.routes.predict import _compute_features
+from ml_model.loader import load_pipeline
+
+
+def stats_employees(session):
+    total = session.query(Employee).count()
+    print(f"\n--- Employés en base ---")
+    print(f"Total : {total}")
+
+    quittes = session.query(Employee).filter(Employee.a_quitte_l_entreprise == True).count()
+    print(f"Ont quitté : {quittes} ({round(quittes / total * 100, 1)}%)")
+
+
+def apercu_employees(session, n=5):
+    rows = session.query(Employee).limit(n).all()
+    print(f"\n--- Aperçu ({n} premiers employés) ---")
+    for e in rows:
+        print(f"  id={e.id} | {e.poste} | {e.statut_marital} | quitte={e.a_quitte_l_entreprise}")
+
+
+def predict_from_db(session, employee_id: int):
+    pipeline, threshold = load_pipeline()
+
+    emp = session.query(Employee).filter(Employee.id == employee_id).first()
+    if not emp:
+        print(f"Employé id={employee_id} introuvable.")
+        return
+
+    data = {col: getattr(emp, col) for col in [
+        "age", "genre", "statut_marital", "poste", "domaine_etude",
+        "niveau_education", "nombre_experiences_precedentes", "annee_experience_totale",
+        "annees_dans_l_entreprise", "annees_dans_le_poste_actuel",
+        "annees_sous_responsable_actuel", "annees_depuis_la_derniere_promotion",
+        "note_evaluation_actuelle", "note_evaluation_precedente",
+        "augmentation_salaire_precedente", "nb_formations_suivies",
+        "nombre_participation_pee", "satisfaction_employee_environnement",
+        "satisfaction_employee_nature_travail", "satisfaction_employee_equipe",
+        "satisfaction_employee_equilibre_pro_perso", "heure_supplementaires",
+        "frequence_deplacement", "distance_domicile_travail", "revenu_mensuel",
+    ]}
+
+    df = _compute_features(pd.DataFrame([data]))
+    proba = float(pipeline.predict_proba(df)[0][1])
+    prediction = int(proba >= threshold)
+    label = "Quitte" if prediction == 1 else "Reste"
+
+    print(f"\n--- Prédiction pour l'employé id={employee_id} ---")
+    print(f"  Poste : {emp.poste} | Réel : {'Quitte' if emp.a_quitte_l_entreprise else 'Reste'}")
+    print(f"  Prédit : {label} (probabilité : {round(proba, 4)})")
+
+
+def stats_predictions(session):
+    total = session.query(Prediction).count()
+    print(f"\n--- Prédictions loggées en base ---")
+    print(f"Total : {total}")
+    if total > 0:
+        derniere = session.query(Prediction).order_by(Prediction.id.desc()).first()
+        print(f"Dernière : id={derniere.id} | {derniere.label} | proba={derniere.probabilite} | {derniere.created_at}")
+
+
+if __name__ == "__main__":
+    session = SessionLocal()
+    try:
+        stats_employees(session)
+        apercu_employees(session, n=5)
+        predict_from_db(session, employee_id=1)
+        stats_predictions(session)
+    finally:
+        session.close()