Fix proxy vote receiver reload lifecycle

[BenCodez]

What changed

• Shut down the existing vote receiver before reloading it on both BungeeCord and Velocity, preventing a second bind to the configured Votifier port.
• Shut down the Bungee receiver during plugin disable.
• Accept the documented test proxy subcommand while retaining vote as a compatibility alias.
• On Velocity, initialize RSA keys before starting the receiver, avoid replacing the active receiver until startup succeeds, and report reload failures to the command sender and log.

Root cause

Proxy reload paths created a new VoteReceiver without closing the current listener. The second ServerSocket bind therefore fails with BindException: Address already in use, leaving reload broken. The proxy command implementations also only recognized vote, although the documented command is test; test was silently ignored.

Impact

Proxy administrators can reload VotifierPlus without leaking the listener, and /votifierplus test <player> <service> now triggers a test vote on BungeeCord and Velocity.

Validation

• Reviewed the one-commit diff against master.
• git diff --check passed.
• Maven tests could not run in this environment because mvn is not installed.