Version 0.2.0 release

Release 0.2.0 expands LLM-assisted generation from ODPC fragments into full
ODPS product drafts built from source material such as transcripts, emails, and
briefs.

Highlights

• open-data-products generate --kind odps-product can now turn each file in an
  input folder into a full ODPS YAML product draft.
• --profile minimal remains the default evidence-first behavior, while
  --profile complete-draft can draft required commercial and governance
  sections such as SLA, dataQuality, and pricingPlans.
• --include-components gives advanced control over optional root-level ODPS
  sections, including access, license, payment gateway, SLA, data quality, and
  pricing content.
• --max-source-chars chunks long transcripts and other large source files,
  extracts facts per chunk, and merges those facts before YAML generation.
• ODPS product generation now uses purpose-specific prompts for fact extraction,
  chunk merging, minimal YAML generation, component drafting, assembly, and
  validation repair.
• Generated ODPS YAML is normalized toward the bundled schema before output,
  including schema-facing SLA.declarative, dataQuality.declarative, named
  $ref links, pricing plan shape, and access field casing.
• The generation result includes review notes, drafted component names, evidence
  gaps, validation status, and repair warnings in --json output.
• ODPS document validation now supports schema-shaped YAML mappings directly,
  including declarative SLA and data quality sections.
• New development documentation explains generation internals, ODPS validation,
  data contracts, ODPC catalog building, ODPG graph building, MCP tools, and the
  cross-spec agent surface.