Add input file as parameter, auto-detecting image size

[Cirromulus]

This PR adds a command line parameter for the input image, fixing #2. Will generate the loss images and .obj in the same folder with the same name as the original image.

Somewhat unrelated, I also added a dynamic number of passes, terminating when the loss compared to the previous run converges (arbitrarily when the improvement is only one percent). This allows for more precision, if the algorithm is able to. If this feature is not desired, I can add a separate PR for that and remove it from here.