Pass schema as schema into CsvReader (pola-rs#15254)

When passed in as dtypes, the schema inference is not skipped. That has
the side effect that only the first `n` columns from the passed-in
schema are eventually used (see the `infer_file_schema_inner` method
in the `polars-io/src/csv/utils.rs` file).

With the change `scan_csv` behaves the same as `read_csv` when used with
a schema having more columns that the file header:

```python
with tempfile.NamedTemporaryFile() as f:
    f.write(b"""
 A,B,C
 1,2,3
 4,5,6,7,8
 9,10,11
 """.strip())
    f.seek(0)
    df = pl.read_csv(f.name, schema=dict.fromkeys("ABCDE", pl.String), truncate_ragged_lines=True)
    print(df)
    lf = pl.scan_csv(f.name, schema=dict.fromkeys("ABCDE", pl.String), truncate_ragged_lines=True).collect()
    print(lf)
...
>>> check()
shape: (3, 5)
┌─────┬─────┬─────┬──────┬──────┐
│ A   ┆ B   ┆ C   ┆ D    ┆ E    │
│ --- ┆ --- ┆ --- ┆ ---  ┆ ---  │
│ str ┆ str ┆ str ┆ str  ┆ str  │
╞═════╪═════╪═════╪══════╪══════╡
│ 1   ┆ 2   ┆ 3   ┆ null ┆ null │
│ 4   ┆ 5   ┆ 6   ┆ 7    ┆ 8    │
│ 9   ┆ 10  ┆ 11  ┆ null ┆ null │
└─────┴─────┴─────┴──────┴──────┘
shape: (3, 5)
┌─────┬─────┬─────┬──────┬──────┐
│ A   ┆ B   ┆ C   ┆ D    ┆ E    │
│ --- ┆ --- ┆ --- ┆ ---  ┆ ---  │
│ str ┆ str ┆ str ┆ str  ┆ str  │
╞═════╪═════╪═════╪══════╪══════╡
│ 1   ┆ 2   ┆ 3   ┆ null ┆ null │
│ 4   ┆ 5   ┆ 6   ┆ 7    ┆ 8    │
│ 9   ┆ 10  ┆ 11  ┆ null ┆ null │
└─────┴─────┴─────┴──────┴──────┘
```

crates/polars-lazy/src/physical_plan/executors/scan/csv.rs

@@ -26,7 +26,7 @@ impl CsvExec {
         CsvReader::from_path(&self.path)
             .unwrap()
             .has_header(self.options.has_header)
-            .with_dtypes(Some(self.schema.clone()))
+            .with_schema(Some(self.schema.clone()))
             .with_separator(self.options.separator)
             .with_ignore_errors(self.options.ignore_errors)
             .with_skip_rows(self.options.skip_rows)