ReduceSum LPT issue with different precision (#33162)
We are enabling CoPilot models on OV GPU.
For one of the models, the issue happens at
https://github.com/openvinotoolkit/openvino/blob/71d7d40b18f31beb2cf1db0f26cd212f9bacd45e/src/common/low_precision_transformations/src/reduce_sum.cpp#L74
DqPrecision is set to FP32 which results in the below issue.
<img width="1907" height="1014" alt="ReduceSum-trans"
src="https://github.com/user-attachments/assets/a41f32ae-1bc2-4c29-bb44-5b86811487f8"
/>
<img width="1907" height="1008" alt="reduce-sum-layer"
src="https://github.com/user-attachments/assets/88c6983c-ce3c-4745-8c8f-b7fa595a73a1"
/>
### Tickets:
- *CVS-177907*
---------
Co-authored-by: Vladislav Golubev <vladislav.golubev@intel.com>
