feat(flow): kbagent flow examples + fix drifted flow schema (port get_flow_examples)

[padak]

Part of #390.

Summary

Add kbagent flow examples (port of get_flow_examples) and, at the same time, fix the drifted hand-written flow schema by porting the authoritative flow JSON schemas.

MCP source (keboola-mcp-server @ agent-v1.60.0)

• Examples: src/keboola_mcp_server/tools/flow/tools.py (get_flow_examples) loads static bundled files via importlib.resources:
  • resources/flow_examples/conditional_flow_examples.jsonl — 2 examples (for keboola.flow; string-UUID ids, graph next: [...]).
  • resources/flow_examples/legacy_flow_examples.jsonl — 2 examples (for keboola.orchestrator; integer ids, dependsOn: [...]).
  • Gated: keboola.flow examples require the project conditional_flows feature ('hide-conditional-flows' not in project_features).
• Schemas (sibling): tools/flow/utils.py FLOW_SCHEMAS → static bundled flow-schema.json / conditional-flow-schema.json.

What kbagent already has

• Examples: nothing — no bundled files, no fetch path (there is no endpoint; bundling is the only option).
• Schema is drifted: commands/flow.py _FLOW_SCHEMA is a hand-written triple-quoted string covering only keboola.flow, drifted independently from the MCP server's authoritative JSON schemas, with no orchestrator variant. flow_service.py has no schema/examples methods.
• Both flow types are already modelled: _FLOW_COMPONENT_CHOICES = ["keboola.orchestrator", "keboola.flow"]; flow new/flow update already take --component-id for both.

Proposed command(s)

kbagent flow examples [--component-id keboola.orchestrator|keboola.flow]

Plus: replace _FLOW_SCHEMA with the bundled authoritative flow-schema.json + conditional-flow-schema.json, so flow schema [--component-id …] serves the correct schema per type.

Implementation sketch

• Bundle the two .jsonl example files and the two schema JSON files copied from the MCP server into src/keboola_agent_cli/resources/flow_examples/ (+ schemas), loaded via importlib.resources. Add the data files to packaging (pyproject.toml / hatch include).
• L2 small loader in flow_service.py returning the parsed list / schema per type.
• L1 flow examples thin command (dual output: JSON list of flow-config dicts; human = numbered JSON blocks). Update flow schema to read the bundled schema per --component-id.

Effort: M

Dominated by bundling static resource files + packaging + tests, not logic.

Acceptance criteria

• kbagent flow examples --component-id keboola.flow and …keboola.orchestrator each return their examples.
• flow schema --component-id … serves the authoritative schema for both flow types (orchestrator variant added).
• --json emits structured dicts.
• Packaging includes the new data files (verified in a built wheel).
• Unit test + E2E test.
• Doc-sync surfaces per Epic: Port keboola-mcp-server agent tools into native kbagent commands #390 checklist.

Risks / open questions

• Static-bundle drift: copying examples/schemas forks a second copy that can rot — exactly the failure already present in _FLOW_SCHEMA. Mitigate with a sync note/script; treat examples as illustrative. (No API to fetch them.)
• Feature gate: kbagent has no conditional_flows/project_features plumbing today. Recommendation: ship ungated v1 (examples are static illustrative text; flow new already accepts both types), document that keboola.flow needs the feature. Adding the gate later is optional.

[padak]

⚠️ Overlaps with #387 — reconcile against its merged result before starting

#387 ([AJDA-2813] Conditional flows support, drop orchestrator, breaking 0.56.0) rewrites the exact surface this issue targets (commands/flow.py, services/flow_service.py). Do not start this issue until #387 merges, then reconcile the scope below.

What #387 already does (and supersedes parts of this issue):

1. Drops keboola.orchestrator entirely. → the "add an orchestrator schema variant / serve both flow types" premise here is obsolete. legacy_flow_examples.jsonl is moot — only keboola.flow remains.
2. Fixes the drifted _FLOW_SCHEMA — the schema half of this issue is done, and done better: flow schema [--full --project] fetches the live schema from the stack; the authoritative CF JSON Schema is vendored to resources/conditional-flow-schema.json (pinned SHA); a new services/flow_validation.py + flow validate command are added.
3. Introduces resources/ + importlib.resources + hatchling packaging (incl. a wheel-inclusion check) — the exact bundling infra this issue needed.

What #387 does not do (confirmed: it adds only flow schema and flow validate, no flow examples, no *_examples.jsonl):

• the actual kbagent flow examples command (port of get_flow_examples) is still missing.

Revised scope after #387 merges

• Drop: the schema-fix / orchestrator-variant work (superseded).
• Keep: add kbagent flow examples for keboola.flow only — bundle conditional_flow_examples.jsonl next to [AJDA-2813] Conditional flows support, drop orchestrator (0.57.0) #387's resources/conditional-flow-schema.json, reuse its importlib.resources loader + packaging, add a thin command + a service loader.
• Effort drops from M → ~S.

Title/body of this issue should be updated accordingly once #387 lands.