3rd set of IIO fixes for the 4.0 cycle.

* A double free occured on an error path in due to an event registration issue.
  The fix is the minimal change rather than possibly reworking this area of
  the core to give a more elegant solution (future work).
* A number of drivers were directly accessing indio_dev->buffer->scan_mask
  to identify the currently enabled channel set.  This may not be correct
  if we have additional clients on the push interface.  The correct option
  is indio_dev->active_scan_mask. This is fixed.
* bmc150 had incorrectly specified sampling frequency (a datasheet confusion
  as they are specified in terms of bandwith - e.g. half the sampling
  frequency).
* hmc5843 wasn't setting it's name and hence the name attribute was
  returning an empty string.
* inv_mpu6050 wasn't clearing the locally held timestamp buffer when the
  hardware fifo was reset. Also an inconsistency existed in the interface
  for the scale of the channels.  Magic numbers were written but real ones
  were used for the reads.  Now uses real numbers (i.e. not array indexes)
  for both.
* fix a missing dependency in the dummy driver. Previously shielded from
  the autobuilders by an earlier build error.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
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=9ZzZ
-----END PGP SIGNATURE-----