[MENFORCER-509] ReactorModuleConvergence should better recognize multimodule roots

**[Robert Scholte](https://issues.apache.org/jira/secure/ViewProfile.jspa?name=rfscholte)** opened **[MENFORCER-509](https://issues.apache.org/jira/browse/MENFORCER-509?redirect=false)** and commented

In case there's an aggregator outside of a multimodule project to trigger one or more multimodules, this rule fails.

In the code is a valid question: // After we are sure having consistent version we can simply use the first one?

Most of the time the answer is yes, but not always.

Attached is an example. 

Running `mvn validate` from base-project succeeds, but not from `extended-project`

Also notice that base-project is executed before extended-project, however it doesn't seem to be the first of the sorted projects?!

---

**Attachments:**
- [reactormoduleconvergence.zip](https://issues.apache.org/jira/secure/attachment/13071139/reactormoduleconvergence.zip) (_2.00 kB_)

Reproduce the failure with the attached reactormoduleconvergence example by running `mvn validate` from both base-project and extended-project. Then inspect ReactorModuleConvergence and how it identifies multimodule roots and orders projects. Done means validation succeeds from both project locations and the rule handles the reported aggregator layout consistently.

Written by the indexing model from the issue text.