Switch to a multiprocess client-sever model

[kunalmohan]

• Client and server now run in separate processes. The server process is a daemon.
• Add a separate panic hook for the server process.
• Use ClientToServerMsg and ServerToClientMsg for IPC. ServerInstruction is to be used on the server side only and ClientInstruction is used on the client-side only.
• config_options are sent to the server with NewClient() message only. They are not passed while spawning the server.
• Tests still run in a single process.

[imsnif]

Hey @kunalmohan - good to see this coming along so quickly! I take it detching/attaching is not yet implemented, right? What happens when Zellij is closed? If it's opened more than once, is it connected to the same daemon? If we merge just this change, how will it affect users? Anything else along these lines I'm not thinking about?

[kunalmohan]

@imsnif You are right; sessions detach/attach is not implemented yet. When zellij is closed, the server also exits, so there is no actual need for the server to be a daemon at the moment (I can remove that if that's better, will just have to comment out a couple of lines). Each client has a different server, so if we open 2 instances of zellij, we'll have 4 processes.
The only thing exposed to the users is the --server <socket_path> that spawns a server listening at the socket. The users should not use this manually, as it is used by the client to spawn a server (since client generates the socket file name). There is no additional feature/functionality added or exposed to the user, so they can continue using zellij as before.

The only thing that might be of concern would be- what happens if either the server or the client crashes? I have added the panic hook implementation that ensures (to some extent) that when either of the processes panic, a message is sent to the other one, to exit. This won't work if the panic is in the main thread in either of the processes, which I suspect would rarely happen.