feat(server): make ComfyUI log level configurable

[rickstaa]

This pull request introduces the comfyui-log-level and comfyui-inference-log-level arguments to the server, allowing users to adjust the log level and reduce excessive logging.

Performance Profile

To ensure this does not negativelly affect the performance I ran the profiler in #80 with the https://github.com/yondonfu/comfystream/blob/main/workflows/comfystream/sd15-tensorrt-api.json workflow. The results are shown below.

With changes

Without using the CLI arguments

AVERAGE - CPU: 415.28%, RAM: 14752.30MB, GPU: 52.93%, VRAM: 3548.00MB

With using the Global ComfyUI log CLI argument

AVERAGE - CPU: 423.68%, RAM: 14779.73MB, GPU: 55.25%, VRAM: 3548.00MB

With using the Inference ComfyUI log CLI argument

AVERAGE - CPU: 424.22%, RAM: 14763.52MB, GPU: 54.25%, VRAM: 3548.00MB

Without changes

AVERAGE - CPU: 412.29%, RAM: 14817.95MB, GPU: 55.45%, VRAM: 3548.00MB

[eliteprox]

There is a logging-level parameter that EmbeddedComfyClient accepts that could be used to set the verbosity from ComfyUI. I'd encourage usage of that, but maybe there's other advantages to wrapping a logger around it. See https://github.com/hiddenswitch/ComfyUI/blob/0fbce0a28299384e4f94ab4ff301f7a566dc3461/comfy/cli_args_types.py#L126.

[rickstaa]

@eliteprox I took a quick look at your suggestion, but it seems that the logging level is not correctly applied in ComfyUI for the EmbeddedComfyClient. The log level is set during the import of EmbeddedComfyClient in the _configure_logging function.

I couldn't find any references to the logging_level argument of EmbeddedComfyClient apart from returning it as the verbose value in the configuration. However, this configuration value doesn't appear to be used anywhere in the codebase. As a result, it seems that this setting is currently only applicable to the regular client.