[Quant][Inductor] Use truncate instead of default rounding round when convert float to uint8 (#105109)
**Summary**
When convert float tensor to uint8 data type as `tensor.to(dtype=torch.uint8)`, PyTorch will directly truncate the decimal. Previously, in `convert_float_to_uint8` we use `_mm512_cvtps_epi32` which uses default rounding mode (round to nearest) to convert float to uint8 which doesn't align with the eager mode behavior. Change `_mm512_cvtps_epi32` to `_mm512_cvttps_epi32` to use directly truncate when convert float tensor to uint8.
**Test Plan**
```
python -m pytest test_cpu_repro.py -k test_to_uint8_rounding_method
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/105109
Approved by: https://github.com/jgong5, https://github.com/mingfeima, https://github.com/jerryzh168