Clean up repository before deprecation

[ManuelFay]

Summary

This is part 1/2 of the repository cleanup planned before deprecating ColPali Engine and directing users toward Sentence Transformers.

This PR prepares the final 0.3.18 release by:

• Updating the supported Torch, PEFT, and Pillow dependency ranges for the last time.
• Adding the new optional [plaid] extra for Fast-PLAID.
• Bringing the changelog up to date for releases 0.3.15 through 0.3.18.
• Fixing Ruff formatting issues in the README and model-family documentation.

No deprecation changes are introduced yet. Those will follow separately in part 2/2.

Important merge note

Please do not squash the commits in this PR.

The dependency updates were intentionally kept as individual commits to preserve their history and make the final maintenance changes easy to trace.

Summary part 2

This is part 2/2 of the repository cleanup before deprecating ColPali Engine.

colpali-engine is now deprecated. The repository and package remain available for research, reproducibility, and existing projects, but new projects and production users should migrate to Sentence Transformers.

Sentence Transformers v6 introduces first-class support for ColPali-style models through MultiVectorEncoder. It provides a unified ecosystem for dense bi-encoders and multi-vector late-interaction models across text, images, audio, and video, depending on the model.

The ColPali and ViDoRe teams provided some assistance with bringing the model configurations to the Hugging Face Hub.

Changes

• Add a prominent deprecation notice after the README introduction.
• Direct new and production users to Sentence Transformers v6.
• Explain the unified embedding and retrieval model support.
• Add installation and inference examples using ColPali-family checkpoints.
• Include the colpali-engine → Sentence Transformers API migration table.
• Link to the Sentence Transformers v6 release, multi-vector documentation, and complete migration guide.
• Preserve the original ColPali documentation for research and reproducibility.
• Thank the users, contributors, and researchers who supported the project.

Migration example

from sentence_transformers import MultiVectorEncoder

model = MultiVectorEncoder("vidore/colqwen2-v1.0")

query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(images)
scores = model.similarity(query_embeddings, document_embeddings)

## Validation

- `ruff check --fix`
- `ruff format --check`
- `git diff --check`

[ManuelFay]

@QuentinJGMace, @HuguesSib , @tonywu71 time to say goodbye 🥹

[QuentinJGMace]

This looks good to me, should we call the last PyPi version 1.0 ? :)

[glesperance]

woww.
congrats colpali team!

and thanks for the amazing engine / work.

this was a blast to be early on