I am working with a Polars DataFrame and need to compute a cumulative sum of data values that resets whenever the cumulative sum exceeds a specific threshold, (absolute value). Here is an example of the DataFrame structure and the expected output:

┌───────┬────────────┬──────────┐
│ flag  ┆ data_value ┆ cum_sum  │
│ ---   ┆ ---        ┆ ---      │
│ i64   ┆ i64        ┆ i64      │
╞═══════╪════════════╪══════════╡
│ -1    ┆ 10         ┆ -10      │
│ 1     ┆ 15         ┆ 5        │
│ 1     ┆ 20         ┆ 25       │
│ -1    ┆ 30         ┆ -5       │
│ 1     ┆ 10         ┆ 5        │ 
│ 1     ┆ 30         ┆ 35       │ <-- Reset because abs(35) >= 30
│ -1    ┆ 10         ┆ -10      │ <-- New sum starts after reset
│ 1     ┆ 7          ┆ 3        │
└───────┴────────────┴──────────┘

In this table, flag and data_value are used to calculate cum_sum. The cumulative sum should reset if the absolute value of cum_sum reaches or exceeds 30 in this case.

I am looking for a vectorized way or a function in Polars to handle this conditional reset efficiently without having to loop through the rows, if possible at all.