Update metrics.py

[Khoa-NT]

• Add --eval_dir_name: name of the eval folder
• Add --run_iter: Another option to measure metrics iteratively to save VRAM, instead of storing all images in a list.
• Update README with new args.

[softyoda]

What is the --eval_dir_name ?
What is an eval folder ?

Is it required for --run_iter ?

[Khoa-NT]

What is the --eval_dir_name ?
What is an eval folder ?

The eval folder is generated from the render.py, which is the test folder by default.
You can run evaluation on another folder with eval_dir_name. E.g., train

Is it required for --run_iter ?

You have to set this flag in other to read each image instead of storing all of them in VRAM.