chop output differs between eager and lazy mode

```
df
.lazy()
.group_by('a')
.agg(mean = pl.col('b').mean())
.ps.with_columns(pl.col('mean').ps_chop.chop(breaks=[-0.01,0,0.05,0.15,1],left_closed=False).ps_enum.drop_unused())
.collect()
```
differs from
```
df
.group_by('a')
.agg(mean = pl.col('b').mean())
.ps.with_columns(pl.col('mean').ps_chop.chop(breaks=[-0.01,0,0.05,0.15,1],left_closed=False).ps_enum.drop_unused())
```

Run both snippets with the reported data and compare the eager and lazy results, including the grouped mean and the chopped enum column. Start by tracing the ps_chop and ps_enum operations used by with_columns; done means eager and lazy execution produce the same output for this reproduction.

Written by the indexing model from the issue text.