docs: add Pocock token-efficiency experiment plans

[richardfogaca]

Linked issue

None.

Summary

• Add an isolated overnight plan for measuring Pocock context rollover and deterministic coordination.
• Add a follow-on workflow portfolio plan covering CI, implementation, QA, cross-repository, synthesis, and open-ended work.
• Link both active explorations from the context index.

Why / context

These plans define how to evaluate token-efficiency improvements without mutating the primary Agor instance, adding a second workflow engine, or treating cached input as literal spend. They make the isolation boundary, evidence requirements, resource ceilings, and adoption gates reviewable before experiments begin.

Implementation notes

The documents permit durable changes only to planning material under context/. Any temporary runtime or harness changes must live in an isolated, push-disabled lab and be removed after evidence collection.

Validation / test plan

• Pre-commit Markdown formatting completed successfully.
• git diff --check
• Reviewed the documents for private paths, credentials, and private Knowledge identifiers.

Screenshots / demos

Not applicable; documentation-only change.

Risks / rollout / rollback

Documentation-only. Revert the commit to remove the plans and index entries.

Out of scope / follow-ups

This PR does not implement Agor product changes or execute the proposed experiments.