feat: add scoped memory records and knowledge promotion

[esafwan]

Summary

Implements the first two development slices from the scoped memory / knowledge bridge plan:

1. docs: add scoped memory and knowledge bridge RFC #274 core schema + tools

   • Adds Memory Record DocType as the canonical scoped memory/data layer.
   • Adds Memory Policy DocType for capture/retrieval/projection policy configuration.
   • Adds Desk actions for activating records, queueing knowledge projection, opening the generated Knowledge Input, and removing projection.
   • Adds permission-aware backend APIs/tool handlers for saving, searching, reading, archiving, and promoting memory records.

2. docs: add scoped memory and knowledge bridge RFC #274 → docs: RFC for adapting Agno's knowledge/vectordb architecture into HUF #178 promote memory to knowledge

   • Adds optional promote_to_knowledge on Memory Record.
   • Projects active memory records into the existing Knowledge Input flow as Text inputs.
   • Keeps Memory Record as canonical source of truth; Knowledge Input is only the searchable projection.
   • Queues existing Knowledge Input processing so the current Knowledge Source indexing pipeline handles FTS/vector backends.

Key files

• huf/huf/doctype/memory_record/
• huf/huf/doctype/memory_policy/
• huf/ai/memory_tools.py

Architecture note

This intentionally does not implement Hindsight-style learning yet. It creates the stable canonical layer first, then connects it to the existing Knowledge Source / Knowledge Input path.

This PR follows the order discussed in #274:

Memory/Data Record
→ Promote to Knowledge
→ Knowledge Input
→ existing Knowledge Source indexing

Related:

• docs: add scoped memory and knowledge bridge RFC #274 — scoped memory/data management and knowledge bridge RFC
• docs: RFC for adapting Agno's knowledge/vectordb architecture into HUF #178 — Agno-style indexing/retrieval architecture
• Add Hindsight integration evaluation and recommendation (docs) #225 — Hindsight long-term memory evaluation

Testing

Not run in this environment.

Recommended validation after checkout:

bench --site <site> migrate
bench --site <site> clear-cache
bench build --app huf

Manual checks:

1. Create a Memory Record with status = Active.
2. Set Promote to Knowledge = 1 and choose a Knowledge Source.
3. Click Queue Knowledge Projection.
4. Confirm a Knowledge Input is created and queued/processed.
5. Search the target Knowledge Source for the memory summary.

[esafwan]

Superseded by the unified implementation in PR #282.

All the code and concepts from this PR have been integrated into that branch, alongside fully functioning background learning agents and hybrid search backends.