Make SmoothQuant more General (#2728)

* Make SmoothQuant more General
Summary:
- Added SmoothQuantConfig as a base config and made corresponding changes in other parts of the flow

Test Plan:
- Qwen 3-8B with example.py and unittest
- Additional test plans requirerd

ETC
- Fix typo in README.md for SmoothQuant

* refactor: use predefined ToyLinearModel

* fix incorrect parameters

* add type hint for dataclass

* use Quantization API for more generalized SmoothQuant API

* add PREPARE_FOR_LOADING mode for loading quantized weight

* update example and doc for updated SmoothQuant API

* remove overused/misunderstood parameters

* remove unused variable from SmoothQuant

* update SmoothQuant docs for user guide

* add benchmark comparison: base vs smoothquant

* add benchmark: w4a8-dynamic

* update docs for a4w8 benchmark

* replace Sec/Tokens with Tokens/Sec for metrics

* update docs for SmoothQuant experiment

* fix typo in README

* rename parser: repo to model

* fix incorrect id: w4a8 -> w8a8

* remove args: precision dtype, `torch.compile`

* rename: precision -> precision dtype in benchmark table

* add args: bias

* fix typo: W4A8 -> W8A8

* fix ci after adding is_bias args

* remove dead annotations in args: smoothing_factor

* remove torch.compile from unittests

* refactor: use ToyLinearModel in AWQ

* remove unused test case: dtype, alpha

* refactor: parametrize `base_config`

* add TODO for future update

* update integration test for new SmoothQuant API

* add unittest: sanity check for smoothquant acc

* bugfix: ImportError for `ToyLinearModel`

* revert: smoothquant unit test name

* revert: integration test

* update docs

* update docs

* add skiptest: no cuda case

Author: namgyu-youn