Fix warning for dynamic quantization args

[kylesayrs]

Purpose

• Squelch warning in case where the user instantiates QuantizationArgs with dynamic quantization

from compressed_tensors.quantization import QuantizationArgs
args = QuantizationArgs(dynamic=True)

compressed_tensors/quantization/quant_args.py:226: UserWarning: No observer is used for dynamic quantization, setting to None
  warnings.warn(

Changes

• Use None as observer default. In the model validator, if dynamic, observer is set to None. If not dynamic, None gets replaced by minmax
• Add help text explaining new default behavior

Testing

• Tests which use dynamic quantization such as tests/test_quantization/lifecycle/test_dynamic_lifecycle.py::test_apply_tinyllama_dynamic_activations now no longer raise an error
• See above test script