cli,server: remove the shutdown timeout upon the first SIGTERM

[knz]

Currently when one sends SIGTERM to a crdb server, the server starts a graceful shutdown.
If the graceful shutdown does not complete under a minute, the code in cli/start.go further triggers a hard shutdown.

@tbg suggests that we should remove this timeout:

• task managers [are assumed to send] SIGTERM first and send a SIGKILL if the process doesn’t terminate within whatever timeout the operator has configured for the task manager
• we’d do a hard shutdown after the second signal, like today, so quitting nodes on the command line remains easy.