add unit tests for agent_spec, source_registry, and service_registry

[iftiquar]

Summary

• Add 94 new unit tests covering three previously untested core modules:
  • test_agent_spec.py (41 tests): AgentSpec dataclass, frontmatter parsing (valid, empty, malformed YAML, missing fences), metadata extraction/validation (name, tools), file discovery (list_agent_files), agent selection by name (select_agent_spec), in-memory loading (load_agent_spec_from_text), and slug generation (slugify_agent_name).
  • test_source_registry.py (24 tests): default source registration (links, sso, git, jira, redmine), transitive dependency resolution (sso→links, git→links), secret and config-field aggregation, custom source registration, and error handling.
  • test_service_registry.py (29 tests): default service registration across infrastructure/application/integration categories, ServiceDefinition properties (volume/image/container naming), dependency resolution (auto-enable infrastructure, postgres pulled in by chatbot/grafana), secret aggregation, category filtering, port configuration, and custom registration.

Motivation

The project had only 8 unit test files for 128+ source modules. These three modules are foundational — agent_spec defines the agent system, source_registry manages data source dependencies, and service_registry wires deployment infrastructure — yet had zero test coverage.

Test plan

• All 94 tests pass locally (pytest tests/unit/test_agent_spec.py tests/unit/test_source_registry.py tests/unit/test_service_registry.py -v)
• No regressions in existing test suite
• isort --check and black --check pass on all new files
• No external service dependencies (tests use tmp_path fixtures and fresh registry instances)