[ARM][CPU] Perf heuristic for winograd convolution (onednn) #21151

Job	Run time
Smart_CI	29s
Docker	54s
Build / Build	11m 22s
Conformance (OP)	9m 18s
OpenVINO JS API	0s
Conformance (API)	2m 7s
NVIDIA plugin	0s
dGPU Tests	0s
iGPU Tests	0s
Debian Packages / Debian Packages	1m 42s
Samples / Samples	4m 51s
C++ unit tests / C++ unit tests	1m 8s
JAX Layer Tests / JAX Layer Tests	2m 6s
ONNX Models Tests / ONNX Models tests	19m 17s
ONNX Runtime Integration / ONNX Runtime Integration	6m 40s
CPU functional tests / CPU functional tests	20m 32s
OpenVINO tokenizers extension / OpenVINO tokenizers extension	3m 49s
Python API tests / Python API tests	1m 52s
Python unit tests / Python unit tests	3m 48s
Pytorch Layer Tests / PyTorch Layer Tests	18m 3s
TensorFlow TF Hub Models tests	0s
PyTorch Models Nightly Scope1 tests	0s
PyTorch Models Nightly Scope2 tests	0s
TensorFlow Hugging Face Models tests	0s
TensorFlow Layer Tests / TensorFlow Layer Tests	14m 40s
TensorFlow Models tests / TensorFlow Models tests	13m 56s
Keras 3 OpenVINO Backend Tests / Keras 3 OpenVINO Backend	4m 35s
PyTorch Models tests / PyTorch Models tests	33m 42s
JAX/Flax Models tests / JAX/Flax Models tests	1m 49s
ci/gha_overall_status	1s
	2h 56m 41s

Provide feedback