fix(molrecbench_wild): give the converter its required track column

[Anai-Guo]

The bug

MolRecBenchWildDataset.evaluate() cannot run at all. It builds a two-column
frame and then calls the converter with a track= keyword:

selected = pd.DataFrame({
    'index': sample_ids,
    'prediction': [prediction_map[index] for index in selected_indices],
})

from .utils.molrecbench_wild import convert_dataframe, score_records

converted, _ = convert_dataframe(selected, track=self.track)

But convert_dataframe takes no track argument — it reads the track from a
required column:

def convert_dataframe(dataframe, *, sheet="Sheet1", id_suffix=""):
    required_columns = {"index", "track", "prediction"}

So every run of any of the three MolRecBench-Wild tracks raises:

TypeError: convert_dataframe() got an unexpected keyword argument 'track'

and simply deleting the keyword is not enough, because the frame that
evaluate() builds has no track column either:

WorkbookConversionError: sheet 'Sheet1' is missing columns: track

(score_records(ground_truth, converted, self.track, ...) a few lines below is
correct — track really is a positional parameter there, which is likely
where the mix-up came from.)

The fix

Supply the track as the column the converter asks for, and drop the keyword.

Verification

Ran the real vlmeval/dataset/utils/molrecbench_wild package at main
(5a91183) against the exact frame evaluate() constructs.

Before — both failure modes:

--- convert_dataframe(selected, track=self.track) ---
TypeError: convert_dataframe() got an unexpected keyword argument 'track'

--- keyword removed but column still absent ---
WorkbookConversionError: sheet 'Sheet1' is missing columns: track

After — records convert, no diagnostics:

records= 2 issues= 0
  id= mol_0001.png smiles= 'CCO'
  id= mol_0002.png smiles= 'c1ccccc1'

(second prediction was fenced ```json output, to check the normal
extraction path still runs rather than falling back to empty records).

pre-commit's flake8 (7.1.2, --max-line-length=120 --ignore=W503) and isort
6.0.1 are clean on the changed file, and yapf produces the same diff before and
after the change — the added lines are untouched by it.

🤖 Generated with Claude Code