Close a dropped TCP session without blocking on its task

[IntellyCode]

IpStackTcpStream::drop waits for the session's task through block_in_place and block_on. block_in_place works by handing the worker's core to the blocking pool, but the pool stops accepting work as soon as the runtime starts shutting down. A session dropped from that point on blocks its worker forever — the core is never handed off, so the task being waited for has no thread left to run on, and the runtime's shutdown is itself waiting for that worker. For a TUN proxy this happens on every exit with open connections: a handful of sessions usually gets away with it, a few hundred hangs the process for good. On a current-thread runtime the same line panics instead, which aborts the process from inside Drop.

This keeps what the waiting was for and drops the waiting. send_fin_n_change_state_to_fin_wait1 is synchronous and writes to an unbounded channel, so the farewell packet is sent from the drop itself, under the same state guard. The exit signal becomes try_send, and the task is aborted rather than awaited — which drops its destroy_messenger, and since the watcher does rx.await.ok(), the session still leaves the stack.