Allow agents to propose semantic enrichment through governed drafts

[ElliotSun]

User Story

As a data steward,
I want AI agents to propose descriptions, tags, classifications, and related semantic enrichments through drafts,
so that AI can accelerate documentation without becoming the source of truth.

Business Value

AI can be useful for:

• description suggestions
• business-name suggestions
• glossary alignment
• tags/classification
• examples
• documentation enrichment

but every proposal should remain reviewable and governed.

Scope

Expose agent-safe semantic proposal operations over M4 drafts.

Tasks

• Add semantic proposal API.
• Require draft context.
• Record proposing actor.
• Track proposed before/after values.
• Analyze proposal through GovernanceDecision.
• Support human review.
• Add tests.

Acceptance Criteria

• Agent-generated enrichment is stored only in draft/proposal state until approved.
• Existing approved semantics remain untouched.
• Agent identity/source is retained.
• Proposal can be reviewed like human-generated changes.
• Rejected proposals do not alter canonical state.

Dependencies

M4
M5 review workflow
US-M7-009

Out of Scope

• automatically accepting LLM output
• generating technical schema
• autonomous publishing