fix(auto): skip blank lines in wav.scp and jsonl file lists

[Lesereingrape]

Summary

prepare_data_iterator() reads every line of a file list without checking whether it is empty, so a blank or whitespace-only line in wav.scp / text.txt raises IndexError (indexing lines[0] on an empty split() result), and a blank line in a .jsonl list raises json.decoder.JSONDecodeError from json.loads(""). Both come from the same missing guard, so skipping blank lines at the top of the loop fixes the two branches with a single two-line change.

Fixes #3757

Reproduction (before):

from funasr.auto.auto_model import prepare_data_iterator
open("wav.scp", "w", encoding="utf-8").write("utt1 /a.wav\n\nutt2 /b.wav\n")
prepare_data_iterator("wav.scp")   # IndexError: list index out of range
# after: (['utt1', 'utt2'], ['/a.wav', '/b.wav'])

Type of change

• Bug fix
• Documentation
• Example or demo
• Runtime or deployment
• Benchmark or evaluation
• Model/training change

Validation

• New test_scp_filelist_skips_blank_lines and test_jsonl_filelist_skips_blank_lines in tests/test_pcm_input_format.py: 2 failed on main at 66d7a4c2 (IndexError and JSONDecodeError), and the whole file passes with the fix — python -m pytest -q tests/test_pcm_input_format.py → 41 passed.

• python -m compileall funasr examples tests → exit 0.

• tests/test_pcm_input_format.py is already in the "Validate NumPy compatibility" workflow run list and funasr/auto/auto_model.py is in its paths filter, so the new tests are exercised by that job on main.

• python -m compileall funasr examples tests

• Docs or links checked

• Runtime/deployment command tested

User impact

Anyone running batch inference through a wav.scp, text.txt or file.jsonl list — the documented entry point for model.generate(input=<list>) — can keep hand-edited or concatenated lists with separator lines instead of the whole batch aborting on one empty line.

Notes for reviewers

• Real entries are still parsed exactly as before; keys and data order are unchanged, and non-matching file extensions are untouched.
• This is the practical limitation to note: a fork PR in this repository does not trigger workflow runs, so the evidence above is local, and the reproduction needs no model weights or downloads.
• Prepared with an AI coding agent; a human maintainer at the PR author's side has not re-reviewed the diff beyond the checks listed above.