WIP: [GPU] Use FP32 accumulator for QK multiplication for 2nd+ token calculation in PagedAttention #18555

Job	Run time
Smart_CI	23s
Docker	53s
Build / Build	12m 14s
OpenVINO JS API	0s
Samples	0s
Conformance	0s
ONNX Models Tests	0s
ONNX Runtime Integration	0s
CPU functional tests	0s
OpenVINO tokenizers extension	0s
iGPU Tests	0s
dGPU Tests	0s
NVIDIA plugin	0s
Debian Packages / Debian Packages	1m 57s
C++ unit tests / C++ unit tests	52s
Pytorch Layer Tests / PyTorch Layer Tests	3m 13s
JAX Layer Tests / JAX Layer Tests	1m 31s
Python unit tests / Python unit tests	2m 41s
Python API tests / Python API tests	0s
TensorFlow Layer Tests	0s
TensorFlow Models tests	0s
PyTorch Models Nightly Scope1 tests	0s
TensorFlow Hugging Face Models tests	0s
TensorFlow TF Hub Models tests	0s
JAX/Flax Models tests	0s
PyTorch Models tests	0s
PyTorch Models Nightly Scope2 tests	0s
ci/gha_overall_status	1s
	23m 45s

Provide feedback