#2 JNI native helper brokenness

I've discovered that the way you do your JNI helpers is apparently also somewhat broken. The OS_Helper.c code doesn't correctly include unistd.h on Linux for the crypt() call (or define __USE_XOPEN to get crypt() prototyped), so the calling conventions are wrong, causing a SEGV when it attempts to process the password lookup. Also, if you connect as a user once (for example, to get the volume list), if the process is run as root for shadow access, the whole process gets setuid()'d to the plain user's UID, so any further shadow auth attempts fail.