Eliminate duplicate messages

Think about whether do we need to eliminate duplicate messages. In the current system given a graph, if one node is dirty then all the paths reaching to that node is reported as dirty.

Start by tracing the graph analysis and dirty-path reporting entry points in the repository to see where messages are produced. Clarify whether identical messages should be emitted once per node or once per path, then define completion as duplicate reports being removed without losing distinct breaking-change information.

Written by the indexing model from the issue text.