Add TIMM integration to extract embeddings

[dnth]

This PR:

• Adds the TIMM integration to fastdup to let users extract embeddings using any models on TIMM.
• Add tests for the TIMM integration.
• Add sample images to run the tests on.

Parameters for TimmEncoder

• model_name (str): The name of the model architecture to use.
• num_classes (int): The number of classes for the model. Use num_features=0 to exclude the last layer. Default: 0.
• pretrained (bool): Whether to load pretrained weights. Default: True.
• device (str): Which device to load the model on. Choices: "cuda" or "cpu". Default: None.
• torch_compile (bool): Whether to use torch.compile to optimize model. Default False.

Usage

import fastdup
from fastdup.embeddings.timm import TimmEncoder

# Compute embeddings
timm_model = TimmEncoder('resnet18')
timm_model.compute_embeddings("../images")

# Run fastdup
fd = fastdup.create(input_dir=timm_model.img_folder)  
fd.run(annotations=timm_model.file_paths, embeddings=timm_model.embeddings)

Here's a bare minimum Colab notebook to try the integration - https://colab.research.google.com/drive/1hDI8SNQU1lhp6d3Q03BhxDk8ILT7dxGl?usp=sharing

I will make another PR for a proper notebook once this integration is merged.

[dnth]

@dbickson @amiralush This PR is ready for review

[dnth]

Merged manually in to version 1.46