feat: add Dakera to Memory Systems

[ferhimedamine]

Add Dakera — Persistent Decay-Weighted Vector Memory for AI Agents

Dakera is a production-grade AI agent memory server built in Rust, with an MCP server providing 83 tools.

• Repo: https://github.com/Dakera-AI/dakera-mcp
• MCP tools: 83 across 15 categories
• Language: Rust

What makes it unique:

• Decay-weighted importance scoring — memories fade like human memory, most-relevant surfaces first
• Hybrid retrieval: BM25 full-text + ONNX vector embeddings
• Knowledge graph construction, entity extraction, cross-agent memory sharing
• Session-scoped memory isolation
• Self-hosted via Docker or cargo install dakera-mcp

Changes:

1. Added to data/projects.json with category memory
2. Added to README Memory Systems section (manually generated, consistent with AUTOGEN format)

Meets criteria: actively maintained (commits within last 6 months), open source, documented README, and novel in its Rust-native decay-weighted approach.