Feature Request: Merge multiple local datasets into one

There's no way to combine two recorded datasets into one. If you recorded the same task in separate sessions (or on separate days), you're stuck with separate `repo_id`s — you can't train on them as a single dataset without merging manually.

`lerobot.datasets.dataset_tools.merge_datasets()` already does this on the lerobot side (concatenates videos + data files into a new dataset). LeLab just doesn't expose it anywhere.

Would like a "Merge" action on the dataset browser: pick two or more local datasets, give the merged result a name, and it writes a new dataset to the local cache.

Start in the dataset browser and trace how local datasets are selected, named, and written to the cache. Read lerobot.datasets.dataset_tools.merge_datasets() to understand its inputs and outputs; done means a Merge action accepts two or more local datasets, asks for a name, and creates the merged dataset locally.

Written by the indexing model from the issue text.