Founder-only: switch a team's superadmin controls.

• review_required: while on, the team's finished investigations park at pending_review (with a Slack ping to the founders) and the output is released to the customer only after a founder approves it.
• hespergpt_enabled: while off, the workspace's HesperGPT claim chat is hidden and the chat endpoints refuse.
• proceed_without_documents: while on, a claim missing a required document still runs to a verdict instead of parking at awaiting_documents, and the claimant is not chased for the document.
• clinical_agents_api: while on, the clinical-reasoning steps run on the hosted OpenAI Agents API with the claim's full documents mounted in a container (fetched from S3 by presigned URL) instead of a truncated OCR-text prompt.
• flag_threshold: the risk score at or above which the team's cases count as Flagged on the Claims board (default 70, range 40-100).
• contents_enabled: while on, the workspace gets Contents (inventory & pricing); while off, /api/contents/* answers 404 for it.

Kept out of TeamSettingsRequest on purpose — customers must not be able to toggle their own flags.