feat: GBrain shadow operating layer integration (v1)

[100yenadmin]

Summary

Implement GBrain as a shadow operating layer in the cortex plugin path, with safe non-authoritative integration points.

Scope

• Keep Cortex and LCM authoritative
• Add GBrain observe-only hooks at safe seams
• Do not affect hot-path recall behavior yet
• Prepare for documentation-side rollout to customers later

Requirements

• Map integration at plugin seam around register(api) -> new CortexClient(...)
• Shadow-mode only, no forced writes into live memory surfaces
• Feature flags / kill switch for all GBrain shadow behavior
• Separate config for GBrain URL/API/model/base URL
• No identity/heartbeat file generation in v1

Acceptance criteria

• Plugin can run with GBrain shadow layer enabled or disabled
• No change to authoritative recall behavior when enabled
• Integration points documented
• Ready for parity comparison logging

References

• gbrain-operating-layer-research-2026-04-15
• gbrain-cortex-plugin-integration-map-2026-04-15
• gbrain-minimax-provider-map-2026-04-15