CSE optimization overwrites struct field name

[cmdlineluser]

Checks

• I have checked that this issue has not already been reported.
• I have confirmed this bug exists on the latest version of Polars.

Reproducible example

import polars as pl

df = pl.LazyFrame({"x": [1, 2], "y": [4, 5]})

(df
  .with_columns(
      q1 = pl.struct(pl.col.x - pl.col.y.mean()),
      q2 = pl.struct(pl.col.x - pl.col.y.mean().over("y"))
  )
  .with_columns(
      pl.col("q1").map_batches(lambda x: [print(f"{x.struct.fields=}"), x][-1])
  )
  .collect()
)

Log output

No response

Issue description

The CSE temp column name ends up as the struct field name - instead of the original LHS expression.

x.struct.fields=['__POLARS_CSER_0x9926168f0d727f4b']

With .collect(comm_subexpr_elim=False)

x.struct.fields=['x']

Expected behavior

Correct field name.

Installed versions

main

[ritchie46]

map_batches is a red herring:

df = pl.LazyFrame({"x": [1, 2], "y": [4, 5]})

(df
  .with_columns(
      q1 = pl.struct(pl.col.x - pl.col.y.mean()),
      q2 = pl.struct(pl.col.x - pl.col.y.mean().over("y"))
  )
  .collect()
).to_dict(as_series=False)

{'x': [1, 2],
 'y': [4, 5],
 'q1': [{'__POLARS_CSER_0x8af40ccf78f1851d': -3.5},
  {'__POLARS_CSER_0x8af40ccf78f1851d': -2.5}],
 'q2': [{'__POLARS_CSER_0x8af40ccf78f1851d': -3.0},
  {'__POLARS_CSER_0x8af40ccf78f1851d': -3.0}]}