[SW-235047] use w8a8 path for per_channel for performance regression fixing (#1629)

https://jira.habana-labs.com/browse/SW-235047

## Essential Elements of an Effective PR Description Checklist
- [X] The purpose of the PR, such as "Fix some issue (link existing
issues this PR will resolve)".
- [ ] The test plan, such as providing test command.
- [ ] The test results, such as pasting the results comparison before
and after, or e2e results


## Purpose

Previous, any HPU fp8 linear will go `hpu_ops.apply_fp8_linear_hpu`,
which is not necessary for per_channel scaling since W8A8 also supports
HPU; which introduced a performance regression for WOQ model.

This PR is to skip per_channel scaling fp8 support in
`hpu_ops.apply_fp8_linear_hpu`


## Test Plan

## Test Result

<!--- pyml disable-next-line no-emphasis-as-heading -->

Signed-off-by: Chendi.Xue <[email protected]>

Author: xuechendi

vllm/model_executor/layers/quantization/fp8.py

@@ -473,7 +473,7 @@ def apply(self,
                 use_aiter_and_is_supported=self.use_aiter_and_is_supported,
             )
 
-        if current_platform.is_hpu():
+        if self.block_quant and current_platform.is_hpu():
             if layer.weight_scale.dim() > 1:
                 weight_scale = layer.weight_scale.transpose(0, 1)
             else: