Only run the CAN logger with RT priority when not polling

We are seeing the whole robot pause for a moment and recover.  I suspect
this is because the CAN logger is running as RT priority and causing
load/timing violations somewhere.  Stopping the CAN logger fixes it.
Instead of fully disabling it, let's try non-rt priority first.

Change-Id: I8d63587bdb6ba067323c318e9d84711b1b1ca638
Signed-off-by: Austin Schuh <austin.linux@gmail.com>
diff --git a/frc971/can_logger/can_logger.cc b/frc971/can_logger/can_logger.cc
index c9e99f7..380410f 100644
--- a/frc971/can_logger/can_logger.cc
+++ b/frc971/can_logger/can_logger.cc
@@ -12,8 +12,10 @@
     : shm_event_loop_(event_loop),
       fd_(socket(PF_CAN, SOCK_RAW | SOCK_NONBLOCK, CAN_RAW)),
       frames_sender_(shm_event_loop_->MakeSender<CanFrame>(channel_name)) {
-  // TOOD(max): Figure out a proper priority
-  shm_event_loop_->SetRuntimeRealtimePriority(10);
+  // TODO(max): Figure out a proper priority
+  if (!FLAGS_poll) {
+    shm_event_loop_->SetRuntimeRealtimePriority(10);
+  }
   struct ifreq ifr;
   strcpy(ifr.ifr_name, interface_name.data());
   PCHECK(ioctl(fd_.get(), SIOCGIFINDEX, &ifr) == 0)