Add a way to kill a REPL buffer without killing the remote kernel

[JonatanSahar]

Hi, thanks for your awesome work! I use it daily :)
I often have a situation where the connection to the remote kernel is cut for some reason or other, leaving a "zombie" REPL buffer. When I kill that buffer, oftentimes after connectivity has been regained, it asks me whether to kill the remote kernel - to which I have to respond "yes" in order to complete killing the buffer, but this kills the remote kernel even though the buffer is not actively connected to it (i.e., commands cannot be run through it).
I then reconnect to the remote server, creating a REPL(2) buffer or something along those lines, and re-associate all code buffers with the new REPL, which is messy.