[quant] allow components_to_quantize to be a non-list for single components

[sayakpaul]

What does this PR do?

Just a small QoL improvement for pipeline-level quantization.

PipelineQuantizationConfig(
    quant_backend="bitsandbytes_4bit",
    quant_kwargs={"load_in_4bit": True, "bnb_4bit_quant_type": "nf4", "bnb_4bit_compute_dtype": torch.bfloat16},
    components_to_quantize="transformer", # instead of ["transformer"]
)

[SunMarc]

Thanks !

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[stevhliu]

Will update the docs with this change once merged! 🫡