Feature/workflow-aware reasoning + shared confidence scoring #1441

Build requirements

• Add a feature/workflow hypothesis layer so the agent can name a likely initiating feature/action.
• Add a lightweight configuration mechanism (file-based is OK) for:
  • endpoint → feature tags
  • feature → service mapping
  • optional operator hints (recently shipped feature, scheduled workflows)
• Produce shared confidence scoring across:
  • correlation evidence
  • topology adjacency evidence
  • feature/workflow hypothesis evidence

Acceptance criteria

• The “periodic feature workflow” synthetic scenario (Issue 4 Revert "Codebase refactoring and cleanup " #2) yields a top candidate that is feature/workflow-labeled with a clear rationale.
• The correlation pathway (Issue 5) includes confidence + what evidence drove it.
• Add unit tests that operator hints change rankings in a controlled fixture.