[feat] Add prediction values to evaluation summary

[hglee98]

Description

This PR aims to add functionality that converts the model's prediction values into a dictionary format and stores them in the predictions.json. In this PR, prediction values have been added for Classification, Segmentation and Detection tasks as follows:

Object Detection

{
    "predictions":
        [
            {
                "bboxes":
                [
                    {
                        "class": int,
                        "name": str,
                        "confidence_score": float,
                        "coords": {
                            "x1": int,
                            "y1": int,
                            "x2": int,
                            "y2": int
                        }
                    },
                    ...
                ]
                "shape":
                {
                    "width": int,
                    "height": int
                }
            }
        ]
}

Classification

{
    "predictions":
        [
            "class": int,
            "name": str,
            "shape": {
                "width": int,
                "height": int
            }
        ]
}

Segmentation

{
    "predictions":
        [
            {
                "segmentation":
                [
                    {
                        "class": int,
                        "name": str,
                        "polygon": list
                    },
                    ...
                ]
                "shape":
                {
                    "width": int,
                    "height": int
                }
            }
        ]
}
    

This feature is not supported on pose-estimation tasks yet.

Related to #599

Change(s)

• Add prediction values for Detection task which includes bounding boxes with class, name, confidence score, and coordinates
• Add prediction values for Classification task which includes class, name, and shape information
• Store predictions in dictionary format in the model summary
• Convert segmentation mask to polygon format when saving predictions artifacts
• No implementation for pose-estimation task in this PR

Code Formatting

If you PR to either master or dev branch, you should follow the code linting process. Please check your code with lint_check.sh in ./scripts directory.
For more information, please read the contribution guide in CONTRIBUTING.md.

Changelog

If your PR is granted to dev branch, codeowner will add a brief summary of the change to the Upcoming Release section of the CHANGELOG.md file including a link to the PR (formatted in markdown) and a link to your github profile.

For example,

- Added a new feature by `@myusername` in [PR 2023](https://github.com/Nota-NetsPresso/netspresso-trainer/pull/2023)

Please enable Allow edits and access to secrets by maintainers so that our maintainers can update the CHANGELOG.md.

[illian01]

Seems predictions are stored in list.
But I think this makes hard to know which is the original input image of the prediction.
How about store it in dict[image_name: prediction]? For example,

{
    "predictions": {
        "test_image.png": {
            "class": int,
            "name": str,
            "shape": {
                "width": int,
                "height": int
            }
        }
    }
}

[illian01]

LGTM!
Please add original image info later.