Patchcore integration - Githubissues

Missions:

[x] Find out the type of the hparams datamember of class BaseAlgo(pl.LightningModule)
- Answer: any type that supports dict-like member accessing. (yes, we can, in fact, send a dict as hparams)
- Meaning: we can drop argparser for our usage of this class and can instead inject our own "hparams" like our model and noiser objects etc...
[x] Filter the important arguments of argparser (which is basically the hparams)
[x] Implement a static (for now at least) method that outputs m copies of noised versions of a single image tensor and then denoising them returning the reconstructed batch.
- Basically copied the method from the jupyter-exp branch at initial_experiment.ipynb
[x] Figure out how to inject our model into BaseAlgo to be used in every test_step like in patchcore.
- Answer: Apparently - as an hparam. lol. We can define our own datamember as self.model = ... but we should consider keeping the model as a value of the hparams dict that is also accessible from everywhere inside BaseAlgo.

I've imported base_algo.py that Elli gave us to our root directory as our next "main.py" after the jupyter notebook.
Added a lot of TODO's to perhaps accelerate the progress of integration of patchcore quickly.