default_transforms() raises unhandled ZeroDivisionError on an empty documents list

[mittalpk]

Describe the bug

ragas.testset.transforms.default.default_transforms() computes a document-length histogram and divides each bin count by len(documents) before validating that documents is non-empty:

bin_ranges = [(0, 100), (101, 500), (501, float("inf"))]
result = count_doc_length_bins(documents, bin_ranges)
result = {k: v / len(documents) for k, v in result.items()}

If documents is an empty list, this raises a raw ZeroDivisionError: division by zero instead of a clear, actionable message.

This is reachable from the public API: TestsetGenerator.generate_with_langchain_docs() and generate_with_llamaindex_docs() both call default_transforms(documents=..., ...) whenever the caller doesn't pass an explicit transforms argument, so passing an empty documents list (or a list that resolves to empty after upstream filtering) crashes with this opaque error.

Note the function already validates one bad-input case explicitly — "too short" documents raise a clear ValueError:

else:
    raise ValueError(
        "Documents appears to be too short (ie 100 tokens or less). Please provide longer documents."
    )

The empty-list case just isn't covered by that same validation, since the crash happens earlier, during the bin-count division.

Ragas version

Reproduced against current main (298b682).

Reproduction

from ragas.testset.transforms.default import default_transforms

default_transforms(documents=[], llm=my_llm, embedding_model=my_embeddings)
# ZeroDivisionError: division by zero

Expected behavior

A clear ValueError (matching the existing "too short" validation pattern), e.g. "The documents list is empty. Please provide at least one document to generate default transforms for."

I have a fix + regression tests ready and will open a PR shortly.