feat: implement task assignment subsystem with pluggable strategies

[Aureliolo]

Summary

• Task assignment subsystem with three pluggable strategies (Manual, RoleBased, LoadBalanced) behind a TaskAssignmentStrategy protocol
• TaskAssignmentService orchestrator that validates task status and delegates to a pluggable strategy
• Frozen Pydantic models: AssignmentRequest, AssignmentResult, AssignmentCandidate, AgentWorkload with model validators (_validate_collections, _validate_selected_not_in_alternatives)
• TaskAssignmentConfig added to RootConfig schema with strategy, min_score, max_concurrent_tasks_per_agent
• Structured logging via 8 event constants in observability.events.task_assignment
• Documentation updated: DESIGN_SPEC.md (§15.3 project structure), CLAUDE.md (engine package description), README.md (feature bullet)

Test plan

• Unit tests for all 3 strategies (ManualAssignmentStrategy, RoleBasedAssignmentStrategy, LoadBalancedAssignmentStrategy)
• Unit tests for assignment models (validators, frozen immutability, edge cases)
• Unit tests for TaskAssignmentService (strategy resolution, delegation, unknown strategy error)
• Unit tests for TaskAssignmentConfig (defaults, boundaries, NaN/Inf rejection)
• Unit tests for error hierarchy (TaskAssignmentError, NoEligibleAgentError)
• Unit tests for event constants (pin-assertions for all 8 constants)
• Unit tests for STRATEGY_MAP registry (keys, types, immutability)
• Unit tests for protocol conformance (isinstance checks)
• Integration test for multi-agent assignment flow
• Parametrized tests for workload distributions and inactive agent statuses
• All 3804 tests pass, 96.60% coverage

Review coverage

Pre-reviewed by 10 agents (code-reviewer, python-reviewer, pr-test-analyzer, silent-failure-hunter, comment-analyzer, type-design-analyzer, logging-audit, resilience-audit, security-reviewer, docs-consistency). 25 findings triaged and addressed.

🤖 Generated with Claude Code

Closes #26
Closes #30