ns_doctor race condition on startup

Because ns_doctor makes a call to names on other nodes, this causes the nameserver to become aware of global names when it might not previously have been aware, potentially killing the doctor in the middle of its init/1 and bringing down the whole node. Here's what the log entry looks like:

SUPERVISOR REPORT <5631.131.0> 2010-07-19 13:17:45
===============================================================================
Reporting supervisor

Child process
errorContext start_error
reason killed
pid undefined
name ns_doctor
start_function

restart_type permanent
shutdown 10
child_type worker