ARCHIVED [experiment over]: An AI agent built to do Ralph loops - plan mode for planning and ralph mode for implementing.

feat(agent): Agent trait, AgentId, AgentContext, AgentOutcome types master

Implements P1d.AC2.1 and P1d.AC2.3 - core agent types needed for the agentic loop. - Add Agent trait with id(), profile(), run(), and cancel() methods - Define AgentId as String type alias - Create AgentContext struct containing work packages, decisions, handoff notes, AGENTS.md summaries, profile, project path, and graph store reference - Define AgentOutcome enum covering Completed, Blocked, Failed, and TokenBudgetExhausted variants - Add comprehensive tests in tests/agent_types_test.rs verifying: * Agent trait can be implemented * AgentContext can be constructed with all fields * All AgentOutcome variants can be constructed and pattern-matched * Mock agent can successfully run and return outcomes Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>