Fix Dataset.take raising IndexError when taking more elements than available

[vidigoat]

Description

Dataset.take(n) is implemented as self.select(range(n)), so when n is larger than the dataset length it raises IndexError: Index N out of range for dataset of size M instead of returning the whole dataset. This is inconsistent with IterableDataset.take, which clamps and returns all available elements — the two were meant to be aligned (see #6813). Dataset.skip already clamps correctly (range(n, len(self))); only take was affected.

This clamps n to len(self) so take returns the whole dataset when asked for more, matching IterableDataset.take and the "first n elements" semantics documented in the method.

Steps to reproduce (before this change)

from datasets import Dataset
Dataset.from_dict({"a": [0, 1, 2]}).take(5)   # IndexError: Index 4 out of range for dataset of size 3.

IterableDataset.take(5) on the same data returns all 3 rows.

Test plan

• Added test_take to BaseDatasetTest (runs both in-memory and on-disk) covering fewer / exact / more-than-available / zero.
• The n > len case fails on current main (IndexError) and passes with this change; existing behavior for n <= len is unchanged.

AI disclosure

This fix was prepared with the assistance of an AI coding assistant; the bug was reproduced and the behavior verified before submitting.