perf: optimize per-token nvfp4 quantization kernel. (#3237)

<!-- .github/pull_request_template.md -->

## 📌 Description

Optimize the performance of the per-token nvfp4 quantization kernel
introduced by #3027.

1. default block size to 128.
2. default to fast math path. rename `TE_EXACT_FP4` to
`TRTLLM_DISABLE_FP4_QUANT_FAST_MATH` and controlled by environmental
variable.
3. change argument list of `get_sf_out_offset_128x4` and
`get_sf_out_offset_8x4`.

TODOs:
1. optimize low latency cases.

## 🔍 Related Issues

<!-- Link any related issues here -->

## 🚀 Pull Request Checklist

Thank you for contributing to FlashInfer! Before we review your pull
request, please make sure the following items are complete.

### ✅ Pre-commit Checks

- [x] I have installed `pre-commit` by running `pip install pre-commit`
(or used your preferred method).
- [x] I have installed the hooks with `pre-commit install`.
- [x] I have run the hooks manually with `pre-commit run --all-files`
and fixed any reported issues.

> If you are unsure about how to set up `pre-commit`, see [the
pre-commit documentation](https://pre-commit.com/).

## 🧪 Tests

- [x] Tests have been added or updated as needed.
- [x] All tests are passing (`unittest`, etc.).

## Reviewer Notes

<!-- Optional: anything you'd like reviewers to focus on, concerns, etc.
-->


<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->

## Summary by CodeRabbit

* **New Features**
* Added environment variable configuration to disable fast-math
optimization in FP4 quantization, enabling behavior alignment with
alternative implementations.

* **Tests**
* Added test fixture to validate FP4 quantization functionality with
fast-math mode disabled.

<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

Signed-off-by: Siyuan Fu <siyuanf@nvidia.com>
Co-authored-by: Ziang Li <ziangli@umich.edu>