Wrong data in hiera murano hash

[Denis Egorenko]

After downscale operations in cluster, murano hash has wrong data during puppet run.

For example, deploy cluster with 3 controllers and then destroy 2 of them. There are no any updates
in murano_haproxy task, because hiera returns old node information.

This is because base fuel task globals has same base hash name for murano. After second run it will
use old plugin data with old information. Then, after updating plugin.yaml with correct info,
on each puppet run will be returned wrong information, because information from globals.yaml and plugins.yaml will be merged into one hash (because of default hiera hash strategy).

[Denis Egorenko (degorenko)]

https://review.openstack.org/#/c/336583/