trace_call_bpf() no longer disables preemption on its own.
All callers of this function has to do it explicitly.

Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Thomas Gleixner <tglx@linutronix.de>