Handle crop_shape=None in Diffusion Policy

[alexander-soare]

What this does

crop_shape=None is allowed and it indicates that no cropping should be done, but various places in the code don't handle it. This PR addresses that.

How it was tested

I tried to train diffusion policy with default settings but crop_shape=None and no exceptions were raised.

How to checkout & try? (for the reviewer)

python lerobot/scripts/train.py \
    hydra.job.name=test \
    env=pusht \
    env.task=PushT-v0 \
    dataset_repo_id=lerobot/pusht \
    policy=diffusion \
    policy.crop_shape=null \
    training.save_checkpoint=false \
    training.offline_steps=200000 \
    training.save_freq=20000 \
    training.eval_freq=10000 \
    eval.n_episodes=50 \
    eval.batch_size=50 \
    wandb.enable=false \
    wandb.disable_artifact=true \
    device=cuda \
    use_amp=true

---

This change is 

[aliberts]

Reviewed 3 of 3 files at r1, all commit messages.
Reviewable status: Some CI tests are either incomplete or have failed!, all files reviewed, all discussions resolved (waiting on @alexander-soare)