BE: PlayerClock::time_out() ignores a running clock (flag-fall undetected on the mover's turn)

[abdulwaarith0]

Bug

backend/modules/chess/src/time_control.rs — PlayerClock::time_out() checks the raw remaining_time field, which is only decremented on stop():

pub fn time_out(&self) -> bool {
    self.remaining_time.is_zero()
}

While a clock is_running, elapsed time is not reflected in remaining_time (it is applied later in stop()). So a player who sits on a running clock past zero is never reported as flagged. get_real_time_remaining() already accounts for the running clock:

pub fn get_real_time_remaining(&self) -> Duration {
    if self.is_running {
        if let Some(last_move_time) = self.last_move_time {
            return self.remaining_time.saturating_sub(last_move_time.elapsed());
        }
    }
    self.remaining_time
}

This is exactly the case flag-detection must catch: the player on the move letting their clock run out without moving. Game::check_time_out() (in game.rs) relies on time_out(), so it cannot detect that timeout.

Fix

pub fn time_out(&self) -> bool {
    self.get_real_time_remaining().is_zero()
}

This does not break the existing test in tests/time_control_tests.rs (it asserts !time_out() on a stopped clock and time_out() after set_remaining_time(0) — both still hold).

Note (latent / related cleanup)

This is currently a latent defect: game.rs is not declared in chess/src/lib.rs (lib.rs only exposes bitboard, time_control, pgn, rating), so check_time_out() is uncompiled, and no live server code imports PlayerClock yet. But PlayerClock/time_out() are exported public API and will bite whoever wires up the clock. game.rs also mixes diesel while the rest of the crate uses sea_orm — possibly worth a separate cleanup issue.