feat: users can filter and group models in hub

[0xSage]

Problem

• Model hub is a super long scroll
• Users are downloading models that are too big - not sure which models are appropriate for their hardware
• Model hub currently shows diff quantizations separately, takes up even more space
• We're not highlighting our own latest models!

Success Criteria A clear and concise description of what you want to happen.

• I can filter by my hardware, and see max ram requirements
• I can filter models by tag
• I can filter models by other attributes in model.json
• Show only the biggest quantization I can use? (dont wanna see the 2gb one if my ram is 32gb)

Additional context Related to #786