Imdlbenco

- Update features
- For all training scripts (train.py, test.py, test_robust.py), we have added TODO comments in areas where we encourage users to make modifications for their benefit.
- Debug:
- [[debug] Fix collapsing issue in Tensorboard Output tab & resolve iteration failure when Effective Batch Size exceeds test set size.](https://github.com/scu-zjz/IMDLBenCo/commit/fc777d1c95175d8540b85be90376e241f2359e72)
- Revise:
- [Optimize test.py by reordering the testing loop to prioritize model ckpt loading over dataset iteration, mitigating I/O bottlenecks from repeated ckpt loading.](https://github.com/scu-zjz/IMDLBenCo/commit/d1a22eb1b91df21ed29e8fcc3513b45ed69ee507)

Thanks SunnyHaze for implementing these codes.

- Code Refactoring
- Move the implementation of train/test transforms to training/testing scripts to avoid modifying the source code when using custom transforms. You can just modify `train.py` easily now!
- Print transform in the log.
- [debug] custom transform needs to implement `get_transform_init_args_names` function to print.
- Debug
- add *args, **kwargs to the __init__ function of CAT-Net
- For [Mesorch](https://github.com/scu-zjz/Mesorch) model, now it can read `image_size` parameter and change any resolution as you like with IMDL-BenCo framework (train.py)
- Documentation Update
- Clarify the relationship between `BalancedDataset` and the other two datasets: [Link](http://localhost:8080/IMDLBenCo-doc/guide/quickstart/0_dataprepare.html#important)

Thanks, the author of WeChat public account `CV模型架构`, for pointing out some of the issues. Thanks SunnyHaze, implement these codes.

- debug for training script:
- Resolved an issue where, when training/testing was initiated by calling train.py/test.py through a shell script, if the parameters in the shell were required by both the model's __init__() and the training script (such as the `--image_size` parameter for models requiring custom input dimensions), the model.__init__() was unable to properly retrieve the parameters. Thanks the contribution of Inkyl !
- debug for implementation of Mantra-Net and SPAN, clear the corresponding training/testing shell script. details reference 72 , thanks the contribution of guojiachen2003 !

Repository itself
- Add detailed bug report for dataset loader.
https://github.com/scu-zjz/IMDLBenCo/blob/76ca78e412ab523a98049ce4c3ecae0e1da739c9/IMDLBenCo/datasets/iml_datasets.py#L59
- Add modular test with pytest.

Develop update
- We will use release to give record for each version after v0.1.22.
- CICD integrated in IMDLBenCo, with automatic pytest path under `./tests/pytest`. We looking forward to your PR contribute to the modular test to IMDLBenCo.
- [Codecov](https://app.codecov.io/github/scu-zjz/IMDLBenCo) visualization badge for code coverage added to Readme.md.