fix openvino panic when run CPU inference on XEON HBM platform #17207

Job	Run time
Smart_CI	23s
Docker	48s
Build / Build	12m 6s
Conformance (OP)	9m 48s
Conformance (API)	3m 19s
iGPU Tests	0s
NVIDIA plugin	2m 50s
dGPU Tests	0s
Debian Packages / Debian Packages	1m 51s
OpenVINO JS API / OpenVINO JS API	1m 27s
Samples / Samples	4m 56s
CPU functional tests / CPU functional tests	22m 24s
C++ unit tests / C++ unit tests	14m 37s
JAX Layer Tests / JAX Layer Tests	2m 7s
ONNX Models Tests / ONNX Models tests	20m 16s
ONNX Runtime Integration / ONNX Runtime Integration	6m 36s
OpenVINO tokenizers extension / OpenVINO tokenizers extension	2m 21s
Python API tests / Python API tests	2m 33s
Python unit tests / Python unit tests	4m 37s
Pytorch Layer Tests / PyTorch Layer Tests	16m 31s
TensorFlow Hugging Face Models tests	0s
TensorFlow TF Hub Models tests	0s
PyTorch Models Nightly Scope1 tests	0s
PyTorch Models Nightly Scope2 tests	0s
TensorFlow Layer Tests / TensorFlow Layer Tests	13m 53s
TensorFlow Models tests / TensorFlow Models tests	11m 47s
JAX/Flax Models tests / JAX/Flax Models tests	1m 51s
PyTorch Models tests / PyTorch Models tests	34m 47s
ci/gha_overall_status	1s
	3h 11m 49s

Provide feedback