fix(dataset): patches torchvision transforms registration

[fracapuano]

Title

Patches comments from @imstevenpmwork on #3153

Type / Scope

• Type: Chore
• Scope: patch

Summary / Motivation

NA

Related issues

See #3152

What changed

• Taking in @imstevenpmwork comments, mostly related to separation of concerns between LeRobotDataset facade and the newly introduced double Writer/Reader backends.

How was this tested (or how to run locally)

• Ran normal tests

Checklist (required before merge)

• Linting/formatting run (pre-commit run -a)
• All tests pass locally (pytest)
• Documentation updated
• CI is green

Reviewer notes

• Anything the reviewer should focus on (performance, edge-cases, specific files) or general notes.
• Anyone in the community is free to review the PR.

[Copilot]

Pull request overview

This PR refines how image transforms are managed for LeRobotDataset/DatasetReader, aligning with the “separation of concerns” goal from #3152 by moving transform mutation behind explicit reader APIs rather than directly touching private state.

Changes:

• Add DatasetReader.set_image_transforms() / clear_image_transforms() and route initialization through the setter for consistent validation.
• Update LeRobotDataset to validate/store image_transforms in the facade and propagate updates via the reader’s public setter (no direct _image_transforms access).
• Add a regression test ensuring reader-level set/clear updates runtime behavior.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
src/lerobot/datasets/dataset_reader.py	Introduces reader-level transform setter/clear APIs and centralizes validation.
src/lerobot/datasets/lerobot_dataset.py	Validates/stores transforms on the facade and propagates via reader public methods.
tests/datasets/test_dataset_reader.py	Adds coverage for runtime transform updates/clearing on DatasetReader.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.