Perfetto: Avoid potential race condition

Fixes a theoretical race where tracing could be stopped and immediately
started again, and as we PostTask the destruction of the tracing session
in theory the new tracing session could be destroyed rather than
the old one.

R=eseckler@chromium.org

Bug: 900920
Change-Id: I2763100bb0915718cf1f1f0f685172d12fc24600
Reviewed-on: https://chromium-review.googlesource.com/c/1335953
Commit-Queue: Eric Seckler <eseckler@chromium.org>
Reviewed-by: Eric Seckler <eseckler@chromium.org>
Cr-Commit-Position: refs/heads/master@{#608129}

services/tracing/perfetto/perfetto_tracing_coordinator.cc

@@ -54,8 +54,8 @@ class PerfettoTracingCoordinator::TracingSession {
       base::ResetAndReturn(&stop_and_flush_callback_)
           .Run(/*metadata=*/std::move(*metadata));
 
-      base::SequencedTaskRunnerHandle::Get()->PostTask(
-          FROM_HERE, std::move(tracing_over_callback_));
+      std::move(tracing_over_callback_).Run();
+      // |this| is now destroyed.
     }
   }