Fix some data races

[ovaag]

1. Add buffer_mutex to getNumMeasurementsBetween to guard against this method reading data while the imu thread is writing to buffer.
2. I noticed when adding a new barometer manager that the imu preintegration started crashing.
   This was because some random baro measurements would have exact same nanosecond stamp as a given imu measurement because they shard hw clock.
   This should now be better handled in dd1672f
   Removed the two-step mutex acqusition in setPropagationBaseState to reduce complexity. This will be catched up by next imu callback anyway?

[nkhedekar]

Thanks for the PR. I'd fixed this locally but hadn't pushed anywhere. Will check the differences between that and this and report back.

[nkhedekar]

Changes look good, thank you! :)