feat: add episodic memory plugin

[Wanbogang]

Summary

Adds an episodic memory system that allows the robot to store past interactions and recall them via semantic similarity, injecting relevant context into the LLM prompt.

Changes

• src/providers/episodic_memory_provider.py — Core singleton provider. Stores episodes in memory, generates embeddings via OpenAI text-embedding-3-small, retrieves similar episodes using cosine similarity, and persists to JSON flat file at config/memory/.{mode}.episodes.json
• src/backgrounds/plugins/episodic_memory_writer.py — Background plugin that flushes episodes to disk every 30 seconds
• src/inputs/plugins/episodic_memory_input.py — Fuser input plugin that recalls relevant past episodes and injects them into the LLM prompt
• src/hooks/episodic_hook.py — Mode lifecycle hooks that load/flush episodes on mode entry/exit with optional TTS announcement
• src/runtime/cortex.py — Calls write_episode() after each LLM tick
• config/simulation.json5 — Registers EpisodicMemoryInput, EpisodicMemoryWriter, and episodic hooks across all 4 modes (slam, navigation, conversation, guard)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ All tests successful. No failed tests found.

📢 Thoughts on this report? Let us know!