Nano: add an overall inf ppl code

[TheaperDeng]

Description

1. Why the change?

related to #5335

2. User API changes

related to #5335

3. Summary of the change

related to #5335

4. sample output

Here is an example on optimize a self defined CNN based NN on Cifar10 dataset.

The whole process completed in 1min on a i5-12600KF.

('onnxruntime_fp32_noipex', {'latency': 0.030279159545898438, 'accuracy': 0.4448})
('onnxruntime_int8_noipex_integer', {'latency': 0.04363059997558594, 'accuracy': 0.3576})
('onnxruntime_int8_noipex_qlinear', {'latency': 0.05829334259033203, 'accuracy': 0.4444})
('openvino_int8_noipex', {'latency': 0.07152557373046875, 'accuracy': 0.422})
('openvino_fp32_noipex', {'latency': 0.07748603820800781, 'accuracy': 0.4448})
('None_int8_noipex', {'latency': 0.18835067749023438, 'accuracy': 0.4452})
('jit_fp32_noipex', {'latency': 0.21636486053466797, 'accuracy': 0.4448})
('jit_fp32_ipex', {'latency': 0.2200603485107422, 'accuracy': 0.4448})
('None_fp32_ipex', {'latency': 0.35202503204345703, 'accuracy': 0.4448})
('None_fp32_noipex', {'latency': 0.5089044570922852, 'accuracy': 0.4448})