Genai/optimum support streaming output

[zhaohb]

Support chunk streaming mode, mainly to reduce the number of decode calls, thereby improving performance

[zhaohb]

@eaidova @as-suvorov
Could you please review this PR quickly? Thank you very much.

• First, that pr supports streaming mode, and secondly, it supports chunk streaming in streaming mode.
• Chunk streaming reduces the number of decode calls.
• We tested the glm4v-nano model using igpu, chunk streaming can bring about a 30% performance improvement in token rate in streaming mode.

[eaidova]

@zhaohb looks like it breaks some other supported use cases, could you please fix:

[ INFO ] Traceback (most recent call last):
  File "/home/runner/work/openvino.genai/openvino.genai/./tools/llm_bench/benchmark.py", line 226, in main
    iter_data_list, pretrain_time, iter_timestamp = CASE_TO_BENCH[model_args['use_case']](
TypeError: run_image_generation_benchmark() takes 6 positional arguments but 8 were given

[zhaohb]

@zhaohb looks like it breaks some other supported use cases, could you please fix:

[ INFO ] Traceback (most recent call last):
  File "/home/runner/work/openvino.genai/openvino.genai/./tools/llm_bench/benchmark.py", line 226, in main
    iter_data_list, pretrain_time, iter_timestamp = CASE_TO_BENCH[model_args['use_case']](
TypeError: run_image_generation_benchmark() takes 6 positional arguments but 8 were given

@eaidova thank you very much, I had fix it.

[sammysun0711]

build_jenkins