Dqc-toolkit

What's Changed
* Added feature `LLMCurate` to compute LLM-based confidence scores for free-text labels using an instantiated Huggingface model and tokenizer.
* Added utils to enable LLM inference with semi-automated prompt construction from an input pandas dataframe / Huggingface dataset object.
* Restructured the code for `CrossValCurate` utils and tests

What's Changed

Update`pyproject.toml` to resolve dependency issues
- Change version constraint for scikit-learn from **`>=1.3.2`** to **`>=1.3.2,<1.5`**

What's Changed
* Fix datatype mismatch between input labels and `predicted_label` returned by `CrossValCurate.fit_transform` in (https://github.com/sumanthprabhu/DQC-Toolkit/pull/9)

What's Changed
* Add issue templates for bug reports and feature requests
* Fix for spurious column issue with `CrossValCurate.fit_transform` (https://github.com/sumanthprabhu/DQC-Toolkit/pull/7)

What's changed
* Improved `calibration_using_baseline` performance by replacing `minmax` based rescaling with unit norm normalization (https://github.com/sumanthprabhu/DQC-Toolkit/pull/1)
* Added multiple version support in documentation (https://github.com/sumanthprabhu/DQC-Toolkit/pull/2)

- Added CrossValCurate to curate noisy text classification datasets with support for
- TfidfVectorizer, CountVectorizer and SentenceTransformer objects as feature extractors
- Sklearn classifiers which implement 'predict_proba' method as curation model
- Calibration of prediction probabilities of curation model
- Added Util functions to
- Simulate noisy labels in text classification
- Fetch information to assist debugging
- Added corresponding tests and documentation