This patch makes rpc calls that are the result of booting an instance
propagate the new block device mapping format.
The way this is done is by adding a 'legacy_bdm' flag to the conductor
(task) API. Since both the scheduler and the compute service propagate
the block device mapping as part of the request_spec field, a flag named
legacy_bdm_in_spec to indicate that was introduced in their respective
APIs.
Currently block_device_mapping is not used by any of the in-tree filters
in the scheduler, so it is assumed that scheduler's boot functionality
has now transitioned to the new format.
This patch also bumps RPC versions of the task, scheduler and compute
manager classes.
Finally this patch propagates the legacy_bdm_in_spec through the
run_instance callback of the compute manager so that, in case of error -
the instance is re-scheduled with the proper flag set.
blueprint: improve-block-device-handling
Change-Id: I5f25ddd4d586dda91061f065c1796be726b0ede3
7ea4a01a9d