0101. User Story: Persist trained models with simple versioning

[aluczak]

As a ML engineer, I want to store models in a central place with version info, so that I can roll forward or back between model versions.

Acceptance criteria:

• Trained models are saved with version identifiers (e.g., timestamps or semantic versions)
• Models are stored on disk or in Azure Blob Storage with a clear path convention
• Backend can be configured to use a specific model version

---

Metadata

• Type: user story
• Parent: 0100. Feature: Model storage and versioning #605