Fix: lttng perf counter deadlock

Fix: lttng perf counter deadlock

Using the ust_lock() to lazily setup the perf counters introduces
a scenario where this lock is nested within the urcu-bp read-side
lock.

However, the LTTNG_UST_WAIT_QUIESCENT ust command requires that
urcu-bp synchronize_rcu() is performed with the ust_lock() held.

This inter-dependency introduces a deadlock:

Thread A                          Thread B

rcu_read_lock()
                                  ust_lock()
                                  synchronize_rcu() (blocked by rcu
                                                     read-side lock)
ust_lock()   <-- deadlock

Introduce a new lttng_perf_lock to protect the lttng perf context
data structures from concurrent modifications and from fork. This
lock can be nested within the ust_lock, but never the opposite.

This removes the circular locking dependency involving urcu bp.

Fixes: #1202
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>