P1-A: Integrate HierarchyBuilder into ingestion (books→volumes→shelves)

[devlux76]

Parent issue: Part of #89

Problem

The design docs describe a hierarchical memory structure (Books → Volumes → Shelves) built by HierarchyBuilder, but the current ingest pipeline only creates a single Book and never invokes hippocampus/HierarchyBuilder.ts. This means:

• The system never builds Volume/Shelf prototypes.
• Hotpath admission quotas for volume/shelf tiers are never exercised.
• Downstream retrieval and consolidation logic that depends on hierarchy cannot function.

Goal

Integrate the hierarchy builder into ingest so that every ingest call constructs/updates:

• Books (per document)
• Volumes (clusters of books)
• Shelves (clusters of volumes)

Ensure the hierarchy is persisted and that the SalienceEngine promotions include:

• Book medoid → page-tier quota
• Volume prototype → volume-tier quota
• Shelf routing prototype → shelf-tier quota

Acceptance Criteria

• Ingest calls buildHierarchy (or equivalent) using the new pages
• Hierarchy objects are stored in IndexedDB (Book, Volume, Shelf records)
• MetadataStore accessors for volumes/shelves are exercised in tests
• Existing integration tests are updated (or new tests added) to validate hierarchy persistence
• No regression in existing ingest/query behavior

Notes

This work is a prerequisite for the dialectical retrieval pipeline, which expects a multi-tier hotpath to exist.