[psteinroe]

Adds a guide on running the formatter on SQL that tools generate, starting with Supabase declarative schemas and pg-delta. pg-delta only formats its output enough to be readable and doesn't plan to grow into a formatter, so the language server formatter is the one to run after each generation. Only docs and mkdocs.yml change.

• New guide guides/formatting_declarative_schemas.md: run format --write on supabase/schemas and supabase/migrations after supabase db schema declarative sync, generate or db pull --declarative, and check formatting in CI.
• pg-delta loads the schema files through Postgres, so formatting doesn't change the diff. The exception is a language sql body written as a string: Postgres stores it as written and the formatter reformats it by default, so the guide recommends skipFnBodies: true to keep a formatted schema syncing with zero diff. begin atomic bodies are stored parsed and aren't affected.
• The local step doesn't use --changed, because it compares <base>...HEAD and misses files that were just generated and aren't committed yet.
• features/formatting.md links to the guide and lists skipFnBodies in the options table, where it was missing.

🤖 Generated with Claude Code