Polars arg_unique for list column

How can I obtain the (first occurence) indices of unique elements for a column of type list in polars dataframe? I am looking for something similar to arg_unique, but that only exists for pl.Series, such as to be performed over a whole column. I need this to work one level below that, so on every list that is inside the column. Given the dataframe

df = pl.DataFrame({
    "fruits": [["apple", "banana", "apple", "orange"], ["grape", "apple", "grape"], ["kiwi", "mango", "kiwi"]]
})

I expect the output to be

df = pl.DataFrame({
    "fruits": [[0, 1, 3], [0, 1], [0, 1]]
})

.list.eval() can be used as a fallback when there is no specific .list.* method currently implemented.

df.with_columns(
    pl.col("fruits").list.eval(pl.element().arg_unique()).alias("idxs")
)

shape: (3, 2)
┌────────────────────────────────────────┬───────────┐
│ fruits                                 ┆ idxs      │
│ ---                                    ┆ ---       │
│ list[str]                              ┆ list[u32] │
╞════════════════════════════════════════╪═══════════╡
│ ["apple", "banana", "apple", "orange"] ┆ [0, 1, 3] │
│ ["grape", "apple", "grape"]            ┆ [0, 1]    │
│ ["kiwi", "mango", "kiwi"]              ┆ [0, 1]    │
└────────────────────────────────────────┴───────────┘