Implement kbagent catalog generate command

[padak]

Summary

Implement automated catalog generation for KBC Explorer. Currently the catalog.json and orchestrations.json data files are produced through a manual 7-step process -- this is not reproducible, not schedulable, and slow.

Goal

A single CLI command that produces a complete, schema-valid catalog snapshot:

kbagent catalog generate --output kbc-explorer/
kbagent catalog generate --tiers tiers.yaml --output kbc-explorer/
kbagent catalog generate --dry-run
kbagent --json catalog generate

Detailed Plan

See docs/PLAN-catalog.md for the full implementation plan, including:

• Architecture (new commands/catalog.py + services/catalog_service.py)
• Tier configuration (YAML + convention-based auto-detection)
• Data collection pipeline (4 phases: parallel fetch, orchestration detail, aggregation, assembly)
• Job statistics aggregation logic
• Orchestration assembly logic
• CLI interface (catalog generate, catalog validate, catalog diff)
• Schema validation via jsonschema
• New dependencies: jsonschema, pyyaml

Subcommands

Command	Purpose
catalog generate	Fetch data from all projects, aggregate, validate, write output files
catalog validate	Validate existing catalog.json against schema.json
catalog diff	Compare two catalog snapshots and show changes

Acceptance Criteria

1. kbagent catalog generate produces output matching manually generated catalog
2. Output passes jsonschema validation against schema.json
3. Full generation for 27 projects completes in under 60 seconds
4. --dry-run shows expected API call count without making requests
5. Tier auto-detection works for ir-l0-*, ir-l1-*, ir-l2-* aliases
6. Error in one project does not block generation for others