docs: record out-of-scope decisions

[johannschopplich]

Adds .out-of-scope/ with one file per rejected feature concept, so the next request for a LangChain adapter or a CSV reader can link the earlier decision instead of re-arguing it. Each file states the decision, the reason – grounded in the spec or earlier closing comments – and the requests it covers.

• framework-integrations.md – LangChain, agent frameworks, pandas, and an MCP server (Add Model Context Protocol (MCP) Server for TOON Format #31, Agent framework integration #44, feat(langchain): add ToonSerializer and ToonOutputParser #45, An interface with Pandas #56)
• format-conversion.md – CSV, YAML, and XML conversion and batch processing (Add features: type-safe models, streaming, plugins, semantic optimization & batch processing #25, Add type-safe models, streaming, plugins, semantic optimization & batch processing #35, TOON to CSV python lib #38, Add CSV Parsing & Writing Support to toon_format #39)
• json-text-helpers.md – json_indent on decode(), encode_json, and a loads alias for json.loads (JSON indentation option in decode method #10, feat: Add JSON indentation option to decode() method #37, feat: add encode_json and loads helpers for better JSON null support #57)
• native-backend.md – a Rust/PyO3 backend (Using the rust library instead of a pure python implementation #9)

Key abbreviation (#25, #35) changes the decoded data in any implementation, so it's recorded in toon-format/spec's encoder-data-rewrites.md. The pydantic extra (#46) and a JSONEncoder.default-style hook (§3) are explicitly not covered.

[github-actions]

Coverage report

This PR does not seem to contain any modification to coverable code.