Add THT Jinja2 data sources as stack output

This simply stores the data structure in the network_data
and roles_data YAML files provided with the -n and -r
options when deploying the overcloud.

This can be generally useful for troubleshooting.

Also the 'overcloud node extract provision' command rely
on the roles data source. Storing it in the stack means
we can get to the data in case the user missplaced the
file originally used, or in the case where we want to
automate the process for all deployed stacks.

NOTE: The idea is to backport this to the release intended
as the upgrade from release, so that the follow on change
Icc6a7a438e9d0f39d003d1cf8ed84d6fb1d5485a can use it during
upgrade.

Conflicts:
  overcloud.j2.yaml

Related: blueprint network-data-v2-ports
Change-Id: I1efecdcd7afa6af3e6b4b26f4435198836db535f
(cherry picked from commit 0b5b914)

overcloud.j2.yaml

@@ -1454,3 +1454,8 @@ outputs:
   AnsibleHostVarsMap:
     description: Map of Ansible Host variables per role
     value: {get_attr: [AnsibleHostVars, value]}
+  TripleoHeatTemplatesJinja2RenderingDataSources:
+    description: The role_data and the network_data used when rendering the THT Jinja2 templates
+    value:
+      roles_data: {{ roles }}
+      networks_data: {{ networks }}