journal: close unix socket after successful probe

[jeremy-gill]

This is done to ensure unix sockets aren't left dangling after each Enabled() call, which
can lead to eventual process resource exhaustion. Fixes #366.

[lucab]

Good catch, and thanks for the separate bug report too! It is my understanding that the garbage collector would eventually take care of the stale socket, and that most consumers would do the probe only once upfront and then cache the result.
However I totally see how this could hit the limit of open FDs in a tight loop, and closing the socket when done is a good practice anyway.

[lucab]

A couple of minor comments on spurious new lines. When fixing those, can you please squash everything into a single final commit?

[jeremy-gill]

The consumer in question that triggered this issue is zerolog. Zerolog's journald backend writer checks journald status for each write (see here). Sockets quickly pile up. Perhaps there's a better way to ensure the connection is still alive in zerolog.