Implémente système multi-utilisateurs avec classement

- Auth cookie simple (SSO-ready) : login/logout, redirection si non authentifié
- Données isolées par utilisateur dans /data/users/{user_id}/
- Migration automatique des données legacy au premier login
- Page classement avec podium et tableau comparatif (score, précision zéro, conformité)
- Nav mise à jour : Classement, affichage utilisateur courant, déconnexion

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

app/classement.py

@@ -0,0 +1,48 @@
+"""Leaderboard stats: compare all users on overtime precision and directive compliance."""
+from models import list_users
+from stats import compute_all_stats, _compliance_over, _day_compliant
+from calcul import minutes_to_hhmm
+
+
+def compute_classement() -> list[dict]:
+    """Return ranked list of users with their key metrics."""
+    results = []
+
+    for user_id in list_users():
+        s = compute_all_stats(user_id)
+        if not s["weekly_balances"] and s["n_jours"] == 0:
+            continue  # skip empty accounts
+
+        balances = [b["delta_min"] for b in s["weekly_balances"]]
+
+        # Precision: % of weeks where |solde| ≤ 30 min (≈ at zero)
+        if balances:
+            pct_zero = round(100 * sum(1 for b in balances if abs(b) <= 30) / len(balances))
+            mean_abs = round(sum(abs(b) for b in balances) / len(balances))
+        else:
+            pct_zero = None
+            mean_abs = None
+
+        conf_pct = s["comp_globale"]["pct"]
+
+        # Combined score (SSO era: will weight per role/contract)
+        score = round(
+            (pct_zero or 0) * 0.6 + (conf_pct or 0) * 0.4
+        )
+
+        results.append({
+            "user_id": user_id,
+            "display_name": user_id.replace("_", " ").replace("-", " ").title(),
+            "n_semaines": len(balances),
+            "n_jours": s["n_jours"],
+            "pct_zero": pct_zero,           # % weeks ≈ 0 HS — higher = better
+            "mean_abs_min": mean_abs,        # mean abs(solde) in minutes — lower = better
+            "mean_abs_str": minutes_to_hhmm(mean_abs) if mean_abs is not None else "—",
+            "conf_pct": conf_pct,           # directive compliance % — higher = better
+            "score": score,                  # combined 0-100 — higher = better
+            "avg_delta": s["avg_delta"],
+            "avg_delta_min": s["avg_delta_min"],
+        })
+
+    results.sort(key=lambda r: -r["score"])
+    return results