WIP: Remove individual scripts and modularize the pre-processing steps

[ntadimeti]

This PR is very big just from the nature of the change. 1) get_intervals.py, bw2h5.py, peak2bw.py files are now modules. All of them are called inside main.py. 2) Due to the change above, a bunch of new options are exposed through config files and some options that are no longer relevant are removed (files_train, val_files, etc). 3) Some config options like print_freq, eval_freq, debug, transform, clip_grad etc were considered unimportant and removed after brainstorming with @avantikalal 4) --sizes_file is renamed as --genome. It can now take in strings "hg19" or "hg38" for picking up the sizes files automatically. Optionally one can also provide a path to a sizes file. 5) --regions option is added. This is only relevant for denoising, not for training. Using this, one can choose a subset of chromosomes to run denoising on. See help for how it works. 6) All tests are updated to reflect this API. expected_results folder files are updated to reflect the new output folder structure. 7) --label is renamed as --exp_name to reflect that it's the experiment name. 8) out_home, exp_name, gpu, distributed flags are now mandatory. Meaning, users MUST provide these, there are no default values in config. @avantikalal @tijyojwad -- I suggest you to take a look at this PR as soon as possible, as I suspect this to take some time to iterate on and merge

[ntadimeti]

rerun tests

[ntadimeti]

Closing this PR since it has become too big to track the changes. Will open several smaller PRs which contain subset of these changes for easy review and merge