Auto-CPP-Local v008

In this release...

• Re-imagining of folder structures/usage.....\Input, .\Output, .\cache\downloads, .\cache\working. It is also clearing downloads and working, upon the event of starting a new project.
• Multiple model implementation, only the current best, in preparation for deepseek v2.5 lite, that possibly, it will work with, but for now we are going with, DeepSeek-V2-Lite-Chat-Q8_0.gguf, DeepSeek-Coder-V2-Lite-Instruct-Q8_0.gguf, chat will do chat/decriptive tasks while instruct will do the code/processing tasks.
• Image Generation Removed to focus on functioning compitently with text. If users wish to use images in the project, they will have to input them into .\Input.
• Intelligent assignment of models into safe spaces of combined CPU/GPU resources, but it will be 1 GPU only, but with vulkan, ie at worst 1 layer from each model on vram.
• Deepseek was becoming confused between, .\main.py and .\scripts\main.py, so I renamed .\main to .\launch.