columns parameter from polars.read_csv not working on empty csv files

[rhipnox]

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 csv

with open("empty_csv.csv", "w", newline="") as file:
    writer = csv.writer(file)
    field = ["x", "y", "z", "t"]
    writer.writerow(field)

with open("not_empty_csv.csv", "w", newline="") as file:
    writer = csv.writer(file)
    field = ["x", "y", "z", "t"]
    writer.writerow(field)
    writer.writerow(["a", "1", "d"])
    writer.writerow(["b", "2", "e"])
    writer.writerow(["c", "3", "f"])

not_empty_data = pl.read_csv(
    "not_empty_csv.csv",
    columns=["x", "y"],
    new_columns=["nx", "ny"],
    dtypes=[pl.String, pl.Int64],
).with_columns(pl.lit("q").alias("q"))

empty_data = pl.read_csv(
    "empty_csv.csv",
    columns=["x", "y"],
    new_columns=["nx", "ny"],
    dtypes=[pl.String, pl.Int64],
).with_columns(pl.lit("q").alias("q"))

Log output

file < 128 rows, no statistics determined
no. of chunks: 1 processed by: 1 threads.

Issue description

While reading an empty csv file, when attempting filter the columns with the columns parameter, it returns all columns.

Expected behavior

#sample code:
import csv

with open("empty_csv.csv", "w", newline="") as file:
    writer = csv.writer(file)
    field = ["x", "y", "z", "t"]
    writer.writerow(field)

with open("not_empty_csv.csv", "w", newline="") as file:
    writer = csv.writer(file)
    field = ["x", "y", "z", "t"]
    writer.writerow(field)
    writer.writerow(["a", "1", "d"])
    writer.writerow(["b", "2", "e"])
    writer.writerow(["c", "3", "f"])

not_empty_data = pl.read_csv(
    "not_empty_csv.csv",
    columns=["x", "y"],
    new_columns=["nx", "ny"],
    dtypes=[pl.String, pl.Int64],
).with_columns(pl.lit("q").alias("q"))

empty_data = pl.read_csv(
    "empty_csv.csv",
    columns=["x", "y"],
    new_columns=["nx", "ny"],
    dtypes=[pl.String, pl.Int64],
).with_columns(pl.lit("q").alias("q"))


"""not_empty_data returns:
>>> not_empty_data 
shape: (3, 3)
┌─────┬─────┬─────┐
│ nx  ┆ ny  ┆ q   │
│ --- ┆ --- ┆ --- │
│ str ┆ i64 ┆ str │
╞═════╪═════╪═════╡
│ a   ┆ 1   ┆ q   │
│ b   ┆ 2   ┆ q   │
│ c   ┆ 3   ┆ q   │
└─────┴─────┴─────┘

empty_data returns:
>>> empty_data 
shape: (0, 5)
┌─────┬─────┬─────┬─────┬─────┐
│ nx  ┆ ny  ┆ z   ┆ t   ┆ q   │
│ --- ┆ --- ┆ --- ┆ --- ┆ --- │
│ str ┆ i64 ┆ str ┆ str ┆ str │
╞═════╪═════╪═════╪═════╪═════╡
└─────┴─────┴─────┴─────┴─────┘
"""

Installed versions

--------Version info---------
Polars:               0.20.30
Index type:           UInt32
Platform:             Windows-11-10.0.22631-SP0
Python:               3.12.2 (tags/v3.12.2:6abddd9, Feb  6 2024, 21:26:36) [MSC v.1937 64 bit (AMD64)]

[cmdlineluser]

Can reproduce.

There seems to be some problem in general as scan_csv also "fails" to select just a subset of columns:

pl.scan_csv("empty_csv.csv").select("x", "y").collect()
# shape: (0, 4)
# ┌─────┬─────┬─────┬─────┐
# │ x   ┆ y   ┆ z   ┆ t   │ # <- ???
# │ --- ┆ --- ┆ --- ┆ --- │
# │ str ┆ str ┆ str ┆ str │
# ╞═════╪═════╪═════╪═════╡
# └─────┴─────┴─────┴─────┘

[ritchie46]

@nameexhaustion can you take this one?

[nameexhaustion]

Will take a look