Fix unreliable timeout

[eLud]

Fix: Notify caller when channel becomes inactive unexpectedly

Problem

When network connectivity is lost (e.g., 100% packet loss), requests hang indefinitely instead of timing out or returning an error. This is a regression
introduced in v1.1.0.

Potential fix for #361

Root Cause

When network drops, NIO's HTTP/2 layer detects the connection loss and triggers channelInactive before the IdleStateHandler can fire its timeout event.
However, channelInactive only closed the connection without notifying the caller:

func channelInactive(context: ChannelHandlerContext) {
    self.closeConnection()
    context.fireChannelInactive()
    // ❌ onResponse was never called - request hangs forever
}

This worked in v1.0.4 because the retain cycle (fixed in #354) kept handlers alive longer, giving IdleStateHandler time to fire. With the fix, cleanup happens faster and channelInactive wins the race—but doesn't notify the caller.

Solution

1. Added a hasResponded flag to track whether the response callback has been invoked
2. Modified channelInactive to call the response callback with .unavailable error if no response was received yet

  func channelInactive(context: ChannelHandlerContext) {
      let shouldNotify = !self.hasResponded
      self.closeConnection()
      if shouldNotify {
          self.hasResponded = true
          self.onResponse(.init(
              code: .unavailable,
              error: ConnectError(code: .unavailable, message: "Channel became inactive", ...)
          ))
      }
      context.fireChannelInactive()
  }

Changes

• ConnectUnaryChannelHandler.swift: Added hasResponded flag and updated channelInactive
• ConnectStreamChannelHandler.swift: Added hasResponded flag and updated channelInactive

Testing

Tested with Network Link Conditioner at 100% packet loss:

• Before: Request hangs indefinitely
• After: Request returns .unavailable error promptly