Add DeepSeek and Qwen models, update Kalosm/candle

[lucksus]

• Kalosm updated, candle: 0.8.2
• Run all models (embedding / whisper) on GPU now if available
• AI task deletion button in Launcher
• DeepSeek R1 models added to list
• Qwen 2.5 Coder Instruct 7b model added to list
• Update CI to new Rust version to fix build
• Test and make Qwen model the install default