Update documents addressing AQT workflow

[namgyu-youn]

AQT (AffineQuantizedTensor) is an old workflow, but the following documents are still describing AQT. We should update or deprecate documents addressing AQT.

• API Reference Dtype: IMO, this page can be dropped because https://docs.pytorch.org/ao/main/api_ref_quantization.html substitute it. Drop document describing outdated (AQT) workflow #3639

Source: https://github.com/pytorch/ao/blob/main/docs/source/api_ref_dtypes.rst & https://docs.pytorch.org/ao/main/api_ref_dtypes.html

• Static Quantization: We can update it to use Int8StaticActivationInt8WeightConfig or drop this post.

Source: https://github.com/pytorch/ao/blob/main/docs/source/static_quantization.rst & https://docs.pytorch.org/ao/main/static_quantization.html

[jerryzh168]

(1) sounds good to drop dtypes page
(2) I think for static quant, we could add an official API similar to AWQ/SmoothQuant actually (observer should be configurable for each tensor subclass instead of hardcoded like AWQ/SmoothQuant, can use protocol for that as well I think), that should be able to compose with any tensor subclass that supports this, it's probably better to work on that first, it could also simplify the tutorial

[namgyu-youn]

(1) sounds good to drop dtypes page

#3639 for the first one, could you look into it?