Is there any Polars-based optimization that can be applied to the apply-lambda methodology in this post Remove substring from column based on another column ?

In the polars dataframe below, how could we remove the “_sub” substrings based on the value of sub?

import polars as pl

pl.DataFrame(
    {"origin": ["id1_COUNTRY", "id2_NAME"],
     "sub": ["COUNTRY", "NAME"]}
)

shape: (2, 2)
┌─────────────┬─────────┐
│ origin      ┆ sub     │
│ ---         ┆ ---     │
│ str         ┆ str     │
╞═════════════╪═════════╡
│ id1_COUNTRY ┆ COUNTRY │
│ id2_NAME    ┆ NAME    │
└─────────────┴─────────┘