Expose ECN information in Controller::on_ack

[B1gG]

Summary

Add ECN (Explicit Congestion Notification) visibility to the Controller::on_ack callback to enable custom congestion control algorithms to react to per-packet ECN CE marks.

Motivation

Currently, Quinn's Controller trait receives ECN CE information only through on_congestion_event, which is triggered after ECN-based congestion detection. However, custom congestion control algorithms may need finer-grained access to ECN information on a per-ACK basis to:

1. Implement ECN-aware algorithms: Custom controllers could implement sophisticated ECN-based congestion control (e.g., DCTCP-like algorithms, L4S, or experimental protocols)
2. Early congestion detection: React to ECN CE marks before they accumulate into a congestion event
3. Telemetry and monitoring: Track ECN mark rates for debugging and performance analysis
4. Research and experimentation: Enable development of novel congestion control mechanisms that leverage ECN

Proposed API Change

Modify the Controller::on_ack method signature to include ECN information:

fn on_ack(
    &mut self,
    now: Instant,
    sent: Instant,
    bytes: u64,
    app_limited: bool,
    rtt: &RttEstimator,
    ecn: Option<EcnCodepoint>,
) {
    // Default implementation remains unchanged for backward compatibility
}

Parameters

• ecn: An Option<EcnCodepoint> indicating the ECN marking on the acknowledged packet
  • Some(EcnCodepoint::Ce): Packet was marked with Congestion Experienced
  • Some(EcnCodepoint::Ect0) or Some(EcnCodepoint::Ect1): ECN-capable but not congested
  • None: No ECN information available (e.g., peer doesn't support ECN, or network stripped ECN bits)

Example Use Case

impl Controller for CustomController {
    fn on_ack(
        &mut self,
        now: Instant,
        sent: Instant,
        bytes: u64,
        app_limited: bool,
        rtt: &RttEstimator,
        ecn: Option<EcnCodepoint>,
    ) {
        // Track ECN CE marks for telemetry
        if let Some(EcnCodepoint::Ce) = ecn {
            self.ce_marks_received += 1;
            // React immediately to congestion signal
            self.reduce_cwnd_early();
        }
        
        // Continue with normal ACK processing
        self.handle_ack(bytes, rtt);
    }
}

[djc]

The core idea seems fine. Your issue description comes off as LLM-generated -- please don't do that.