Fix Llama3 miner 4bit flag, update requirements, etc

[dbobrenko]

Changes

1. Add autoawq dependency required to run Llama3 70B AWQ model with miner.
2. Upgrade transformers to 4.41.1, which is required by autoawq.
3. Replace HuggingFaceMiner model_kwargs with quant_config, which previously resulted in error:

AttributeError: 'BitsAndBytesConfig' object has no attribute 'get_loading_attributes'

4. Update miner and validator README with Llama3 requirements.
5. Increase neuron.llm_max_allowed_memory_in_gb default value from 60 to 62. Default value 60 results in:

ValueError: No available memory for the cache blocks. Try increasing gpu_memory_utilization when initializing the engine.

[p-ferreira]

LGTM