Fix the notification pipe for squashfuse in case of failure

[ariel-miculas]

The following scenario now correctly sends a failure notification to the pipe:

$ mkfifo /tmp/notify_pipe
$ cat /tmp/notify_pipe

then run squashfuse:

$ ./squashfuse -o notify_pipe=/tmp/notify_pipe /dev/null /mnt

This fix only applies to squashfuse, squashfuse_ll was behaving as expected.

Fixes #112

[DrDaveD]

Please include the expanded automated test you described in #112. I would think that could be made more compact though, at least a loop with starting something like for prog in squashfuse squashfuse_ll that executes the same tests with both $prog.

While you're fixing things, also please break up that long new -o notify_pipe= line in sqfs_usage into two lines so no line is longer than 80 characters.

[ariel-miculas]

Added tests/notify_test.sh which checks both successful and unsuccessful squashfuse(_ll) mounts.

[ariel-miculas]

@DrDaveD could you make a new release once this is merged?

[DrDaveD]

Yes, that looks better. Unfortunately the freebsd check failed, which it does sometimes, and the only way I have found to force that to re-run is to push a commit. So please recreate the current commit and force push again.

[ariel-miculas]

The CI gods have shown mercy this time.