Fixes #25092: Refactor CachedReportingService to make persistance simpler

[fanf]

https://issues.rudder.io/issues/25092

So, it seems that a lot of things are going on here, but that's not so much:

• the first two commits are linked to pr #5754 and #5755 . So there's only the third commit to consider.
• that should not change anything on Rudder behavior (bravo François, great commit, does nothing :facepalm: )

This commit splits the big CachedReportingServiceImpl into four parts:

• a dumb NodeStatusReportRepository that only store/retrieve NodeStatusReports. Also add a simple implementation in memory. Latter, we will be able to persist in a psql backend storage
• a new extremely simple ReportingService using previous NodeStatusReportRepository. With that and the availability of PolicyType, it's now apparent that that service is just syntaxic sugor on top of filtering NodeStatusReports.
• a ComputeNodeStatusReportService that get the logic with the queue but change nothing in it for now. It updates NodeStatusReportRepository with its updates. Latter on, we will be able to add a check on expiration based on node parameter about how long to keep expired compliance.
• a FindNewNodeStatusReports that only has the logic to retrieve information for new NodeStatusReports in base from expected reports and run. The fact that it is now independant from RepositoryService method signature allows to show that RuleId, DirectiveId and PolicyType paramters are never used in it. We will be able to simplify it even further more latter on (his only job should be to retieve RunAndConfigInfo, and let ComputeNodeStatusReportService do the computing).

For now, we keep the same init logic than we had before the refactoring: there's a bootstrap action that expires all node compliance, so they are all computed back which now init the in-memory NodeStatusReportRepository. Of course once we have that in base, we will get it from there, only asking for the missing nodes if any.

Then, the commit adapt signatures where ReportingService was used.

It alos remove ReportingServiceTest which was only comments (very useful) ; and adapt CachedReportingServiceTest to show that things still work when splitted in several pieces. One value change in that test because we don't exactly have the same call chain: now, the call to trigger a computation is explicit, and the chain is smarter on what should be updated (following https://issues.rudder.io/issues/24652 / https://github.com/Normation/rudder/pull/5737)

All in all, the new code is much, much simpler and get ride of a lot of oranically-grown-not-refactored code ; and it will be rather easy to add the missing parts for persisting and not expiring compliance.