Added support for Facebook''s Segment Anything

[Flippchen]

This Pull Request brings a significant update to our project by introducing the SamSession class, specifically designed to seamlessly integrate with Facebook's advanced Segment Anything model (https://github.com/facebookresearch/segment-anything).

The key highlights of this PR include:

SamSession Class: A dedicated class tailored for the Segment Anything model, allowing users to take full advantage of its capabilities.
Edited Remove Function: The addition of a new parameter to the remove function, enabling the input of target points.
Customized Prediction Function: A separate prediction function, distinct from the other classes, to optimize the usage of the Segment Anything model.
Multi-Model Download Support: An improved download function that now supports downloads for multiple models.

@danielgatis Suggestions for improvement are welcome. I will also add the CI support soon.

You can try it with the following code:

from pathlib import Path
from rembg import remove, new_session

here = Path(__file__).parent.resolve()

model_name = "sam"
session = new_session(model_name)
image = Image.open(here / ".." / "examples" / "car-2.jpg")

input_point = np.array([[600,350]])
input_label = np.array([1])

image = remove(image,session=session, input_point=input_point, input_label=input_label)

image.show()

[danielgatis]

wow! it is an amazing contribution thank you so much!