fix FullClusterRestart tests now that 7.0.0 can be an "oldCluster"

[jakelandis]

Now that 7.0.0 can be an "old" cluster the following needs to be updated:

• Needs a refresh() in addition to _flush :Don't refresh on _flush _force_merge and _upgrade #27000
• A single shard is not always the case for the old cluster.
  (greaterThan --> greaterThanOrEqualTo)
• More then 1 shard is explicitly needed to shrink down to 1 shard
• Disallow a type name in the mapping with out a special flag
  (include_type_name) : Update the default for include_type_name to false. #37285
• Only require include_type_name for 6.7, not 7.0.0
• Fix expectations for template mappings from a snapshot.

This commit favors using the non-deprecrated and typeless APIs if the
"old" cluster is 7.0.0 or greater, and thus all of the relevant expected
depreciations are removed.

closes: #41222

[jasontedor]

This looks similar to problems that we faced in 7.x and master already (which makes sense as full-cluster restart there would be running against these versions). Rather than re-do that work, can be we bring back work that was already done on 7.x or master so that these branches look roughly the same?

[elasticmachine]

Pinging @elastic/es-core-infra

[jakelandis]

This looks similar to problems that we faced in 7.x and master already (which makes sense as full-cluster restart there would be running against these versions). Rather than re-do that work, can be we bring back work that was already done on 7.x or master so that these branches look roughly the same?

doh! I should have looked at the newer branches before starting this. A quick glance and the two changes are very similar. I will close this PR and backport the 7.x version.