`get_all_study_summaries` ignores constraints when selecting `best_trial`

Expected behavior

StudySummary.best_trial should match Study.best_trial for constrained studies. Only feasible trials (all constraint values <= 0) should be considered.

If every complete trial is infeasible, best_trial should be None. Study.best_trial already raises ValueError in that case.

Environment

• Optuna version: 5.0.0.dev
• Python version: 3.14.7
• OS: Linux-6.1.177-224.371.amzn2023.x86_64-x86_64-with-glibc2.34

Error messages, stack traces, or logs

study.best_trial: {'x': 5.0} value=5.0 constraints={'c': 0.0}
summary.best_trial: {'x': -10.0} value=-10.0 constraints={'c': 15.0}

Steps to reproduce

1. Run the following:

import optuna


def objective(trial):
    x = trial.suggest_float("x", -10, 10)
    trial.set_constraint("c", 5 - x)  # feasible iff x >= 5
    return x


study = optuna.create_study(direction="minimize")
study.enqueue_trial({"x": -10.0})  # infeasible, better objective
study.enqueue_trial({"x": 5.0})  # feasible
study.optimize(objective, n_trials=2)

print("study.best_trial:", study.best_trial.params, study.best_trial.value, study.best_trial.constraints)

summary = optuna.get_all_study_summaries(study._storage)[0]
print("summary.best_trial:", summary.best_trial.params, summary.best_trial.value, summary.best_trial.constraints)

1. Study.best_trial is the feasible trial (x=5.0). StudySummary.best_trial is the infeasible trial (x=-10.0).

Additional context (optional)

get_all_study_summaries picks best_trial with min/max over complete trials and does not filter infeasible trials. Reproduced with InMemoryStorage, sqlite, and JournalStorage.