refactor(train): remove EMA support from training pipeline

Drop the opt-in EMA-shadow feature entirely: EMAConfig, the `ema` field on
TrainPipelineConfig, all EMA logic in lerobot_train.py (setup/resume, per-step
update, W&B observability, checkpoint save, EMA-model eval, and the sibling
`<repo_id>-ema` hub push), and the ema-pytorch dependency.

Co-authored-by: Cursor <cursoragent@cursor.com>

pyproject.toml

@@ -85,11 +85,6 @@ dependencies = [
     "termcolor>=2.4.0,<4.0.0",
     "tqdm>=4.66.0,<5.0.0",
 
-    # Training utilities
-    # EMA of policy parameters (Diffusion Policy / pi05 style). Tiny
-    # pure-python dependency — preferred over a hand-rolled implementation.
-    "ema-pytorch>=0.7.7,<1.0.0",
-
     # Build tools (required by opencv-python-headless on some platforms)
     "cmake>=3.29.0.1,<4.2.0",
     "setuptools>=71.0.0,<81.0.0",