docs: add SQL RAG schema-context application to README

[mingjerli]

Summary

• Positions clgraph as the context layer for SQL RAG (retrieval-augmented text-to-SQL) systems: clgraph does not do retrieval or embedding, but produces lineage-aware, description-rich, PII-filtered schema context that any retrieval stack can index
• Adds SQL RAG? Schema context is graph export. to the "One graph. Many applications." list
• Adds an SQL RAG context bullet under LLM-Powered Features
• Adds a new README section "SQL RAG: Schema Context for Retrieval Pipelines" after the Text-to-SQL section, with two code examples:
  • Building one retrieval document per table with PII columns filtered out
  • Expanding retrieved context with trace_column_backward() at question time
• Adds a SQL RAG bullet to the Use Cases section

Test plan

• Both new code examples executed against the current API: the PII filter drops email from the analytics.customers document, and the backward trace resolves lifetime_value to raw.customers.lifetime_value
• All APIs referenced (get_columns_by_table, col.pii, trace_column_backward, propagate_all_metadata, generate_all_descriptions, to_json) verified to exist in src/clgraph/pipeline.py / models.py
• make pre-commit hooks passed (ruff format + lint clean; ty diagnostics are pre-existing and non-blocking)

Companion docs-site PR adds a matching "SQL RAG" use case page section.