Atacworks

This is a pre-release candidate for AtacWorks v0.2.0. The main features of the pre-release include :
* Addition of end-to-end tests.
* Addition of setup scripts.
* updating documentation, linting of code.
* Ability to provide multiple input channels for training, inference.
* Introduced config files for setting up training parameters and for defining model structure.

_**Hotfix**_ Fixed an issue with generating regression bigwg files when some intervals had all 0 valued outputs.

This is the first official release of AtacWorks. AtacWorks is a deep learning-based toolkit for denoising and peak calling from noisy ATAC-Seq data. While currently tested only on ATAC-Seq, AtacWorks can also be applied to other epigenomic data types such as ChIP-Seq or DNase-Seq. A detailed description and results for several use cases are given in the preprint: https://www.biorxiv.org/content/10.1101/829481

The main components of this release are:

1. Data reading and writing from BED, BEDGRAPH and BigWig formats
2. Training deep learning models using a customizable resnet architecture
3. Pre-trained models that can be applied to new data
4. Inference using a newly trained or provided model, producing a denoised ATAC-Seq signal and peak calls
5. Evaluating model performance on denoising and peak calling tasks.