FEATURE: Speedup content cache flush by using cte in findAncestorNodeAggregateIds

[mhsdesign]

Followup to #5221

ContentGraph::findParentNodeAggregates becomes slower on bigger datasets. Due to the mass on executions on a cr:replay, this sums up very quickly. Via #5268 the query will be improved but this pr introduces ContentGraph::findAncestorNodeAggregateIds to make this operation as performant as possible.

• Adds comment for CacheFlushingStrategy strategies

• Introduces ContentGraphInterface::findAncestorNodeAggregateIds using native sql cte to speedup cache flushing. (see comment)

• Move test which creates illegal state to content graph package and use native sql to create the state to not run any catchup hooks. Previously we needed to handle the case of infinite loops to not crash:

  Prevent infinite loops
  NOTE: Normally, the content graph cannot contain cycles. However, during the
  testcase "Features/ProjectionIntegrityViolationDetection/AllNodesAreConnectedToARootNodePerSubgraph.feature"
  and in case of bugs, it could have actually cycles.
  The content cache catchup hook leverage this method and would otherwise be hanging up in an endless loop.
  That's why we track the seen NodeAggregateIds to be sure we don't travers them multiple times.

• fixes a bug where you cannot replay because the workspace is "missing" and no content graph exists

Upgrade instructions

Review instructions

Checklist

• Code follows the PSR-2 coding style
• Tests have been created, run and adjusted as needed
• The PR is created against the lowest maintained branch
• Reviewer - PR Title is brief but complete and starts with FEATURE|TASK|BUGFIX
• Reviewer - The first section explains the change briefly for change-logs
• Reviewer - Breaking Changes are marked with !!! and have upgrade-instructions

[nezaniel]

minimal naming stuff, looks fine otherwise (and is a lot faster in the non-trivial test cases)

adjusted :)

[dlubitz]

Looks fine to me by reading. I left two comments.

[bwaidelich]

+1 by reading, thanks!