Missing support for ground truth files where no events occur

[benjaminsshoffman]

If a ground-truth file does not contain any events, we get an error ValueError: ground_truth audio ids do not match audio_ids.

psds1, percls_psds1, _, _ = sed_scores_eval.intersection_based.psds(
                                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/home/representation-learning/.venv/lib/python3.12/site-packages/sed_scores_eval/intersection_based/psds.py", line 89, in psds
    (effective_tp_rate, effective_fp_rate), single_class_psd_rocs = psd_roc(
                                                                    ^^^^^^^^
  File "/mnt/home/representation-learning/.venv/lib/python3.12/site-packages/sed_scores_eval/intersection_based/psds.py", line 192, in psd_roc
    intermediate_statistics, audio_ids = accumulated_intermediate_statistics(
                                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/home/representation-learning/.venv/lib/python3.12/site-packages/sed_scores_eval/intersection_based/intermediate_statistics.py", line 168, in accumulated_intermediate_statistics
    scores, ground_truth, audio_ids = parse_inputs(scores, ground_truth)
                                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/home/representation-learning/.venv/lib/python3.12/site-packages/sed_scores_eval/base_modules/io.py", line 35, in parse_inputs
    ground_truth = parse_ground_truth(ground_truth, tagging=tagging, audio_ids=audio_ids)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/home/representation-learning/.venv/lib/python3.12/site-packages/sed_scores_eval/base_modules/io.py", line 69, in parse_ground_truth
    raise ValueError(

There is a way to pass audio_ids in base_modules.io.parse_ground_truth, but this option isn't accessible from the high-level functions I want to call (e.g. psds), as far as I can tell.

This is worth fixing because in detection settings, sometimes we have files where no events occur, but we still want to avoid false positives in these cases.