v0.2.1

What's Changed

• feat: implement mjai replay and demo ML project by @smly in #76
• feat: refactor arch, add configurable kuikae modes and update ml demos by @smly in #79
• feat: update cql observation handling by @smly in #81
• feat: add online RL with native game logic refinements by @smly in #82
• refactor: improve game state architecture with new structs by @smly in #83
• feat: improve ml baseline and native environment interaction by @smly in #84
• refactor: ML baseline into src-layout package with DQN online training by @smly in #87
• fix: use mjai_to_tid instead of parse_tile_136 for mjai tile parsing by @smly in #88
• feat: optimize worker rollout: batched multi-env inference with torch.compile by @smly in #89

Full Changelog: v0.2.0...v0.2.1