notebooks: Update multi-tenant schema walkthrough for two-part series

[jeremydaly]

Summary

Updates the multitenant_schema_walkthrough.ipynb companion notebook to track the now-published two-part From Prompt to Persistence blog series instead of a single article:

• Part 1 — Designing Multi-Tenant Agent Memory Schemas for SaaS: the durable layer — eight typed memory tables sharing the same scope columns, with tenant isolation enforced by row-level security.
• Part 2 — Putting the Multi-Tenant Agent Memory Schema to Work: retrieval, versioned supersession, and a provable right-to-forget cascade.

The change is documentation and freshness only — no SQL semantics were altered.

What changed

• Intro cell now links to both articles and describes the durable-layer / put-it-to-work split.
• Minor prose tightening in the intro, knowledge-base, RLS, and unified-retrieval sections.
• Re-ran the notebook so execution counts and cell outputs (generated entity IDs, supersession timestamps, cleanup output) reflect a current end-to-end run.

Verification

Every SQL statement was cross-checked against both articles and runs in article order against a live Oracle AI Database Free (26ai) instance: schema DDL, RLS policy, session context, seed data, the canonical retrieval predicate, supersession, the right-to-forget cascade, and the unified retrieval query. The VECTOR(384, FLOAT32) dimension (vs. the articles' illustrative VECTOR(1024, FLOAT32)) matches the ALL_MINILM_L12_V2 embedding model and is called out in the notebook's Prerequisites.

Notes for reviewers

• Requires Oracle AI Database Free at localhost:1521/FREEPDB1 (or DB_DSN), the ALL_MINILM_L12_V2 ONNX model loaded, and a .env with DB_USER / DB_PASSWORD / DB_DSN.
• No dependency or infrastructure changes.