[Treasury] Implement Validation and Error Handling for win extraction

[jwilson4578]

User Story
as a developer supporting the federal strategy data experience, I want to ensure the data received from the win extraction pipeline is correctly formatted and validated before its stored, AI is behaving as expected, pipeline does not re-run duplicate data, and errors occurrences are properly handled

Acceptance Criteria

1. The system does not re-extract wins from source documents
2. The AI cannot invent strategic plan goals, objectives, KPIs, IDs, or source references.
3. AI output is validated before storage.
4. Invalid AI output is logged and does not corrupt stored data.
5. The system does not force alignment based only on broad or generic language.
6. KPI alignment is only assigned when the win plausibly supports or demonstrates progress toward that KPI.

Definition of Done
The win extraction pipeline completes with all errors being logged, AI evaluation behaves per ACs.
When a win is extracted, the system prevents re-extraction of same win, and performs validation to ensure data quality