docs: describe how pg_dump schema snapshots load

[omer-cengel]

Summary

Documents that a pg_dump --schema-only file can be configured directly as a schema snapshot, so services that keep their schema outside SQL migrations (for example a Liquibase changelog or a JPA-generated schema) can use their dump as sqlcj's schema input.

Changes

• Explain that a dump loads unchanged as one schema file, with nothing to edit, remove, or reorder.
• List the dump content sqlcj ignores rather than rejects: psql meta-commands including \restrict/\unrestrict, the SET preamble and set_config call, the identity form pg_dump writes, constraints, indexes, ownership, grants, policies, comments, extensions, domains, composite types, functions, triggers, views, materialized views, and every schema other than public.
• Describe how partitions written as CREATE TABLE plus ATTACH PARTITION are modeled, and when a dump and its migration history can differ.
• List the integration tests that prove a migration history, a hand-written snapshot, and a dump generate identical output.

Scope and non-goals

Documentation only; no change to parsing, analysis, or generated code.