Version 0.1.9 release

Release 0.1.9 adds an LLM-assisted ODPG graph builder for ODPC fragment
folders and tightens selected-kind generation guidance.

Highlights

• open-data-products odpg-build <fragments/> --output <graph.yaml> now builds
  one ODPG graph from ODPC product reference, use case, objective, and signal
  fragments.
• The new graph builder converts ODPC fragments into ODPG nodes
  deterministically, then asks the configured LLM provider to infer only the
  graph edges.
• A dedicated odpg_edges_from_odpc_fragments.md prompt keeps edge inference
  separate from node creation, so generated edges must reference known node ids.
• The public Python API now exposes build_graph() and write_graph() for
  ODPG graph construction from ODPC fragments.
• New course guidance shows the full workflow: generate ODPC fragments, build an
  ODPC catalog, build an ODPG graph from the same fragments, and generate the
  graph explorer.
• Generation examples now prefer type-specific source folders such as
  source_docs/products/, source_docs/use_cases/,
  source_docs/objectives/, and source_docs/signals/ so each explicit
  --kind prompt receives matching source material.
• README, command docs, llms.txt, and guide examples now consistently show
  open-data-products generate with the required concrete --kind value.