fix(lease): should not panic on keep_alive

[Dreamacro]

Currently in LeaseKeepAlive, calling the keep_alive function may cause a panic if a send operation is performed on req_tx.

There may be some network fluctuations due to poor network conditions in my usage environment (due to cross-border use). I have a keep-alive logic that calls keep_alive_for for retries during fluctuations, which caused this panic in actual scenarios.

thread 'tokio-runtime-worker' panicked at 'emit keep alive request to channel: SendError(LeaseKeepAliveRequest { id: 7587869483910042079 })', /home/.cargo/registry/src/index.crates.io-6f17d22bba15001f/etcd-rs-1.0.0/src/lease/mod.rs:72:14

Therefore, when send fails, I use the same logic as keep_alive_for to return a channel close error.

[zarvd]

LGTM