write duration to parquet but read as int64

[Liyixin95]

Describe the bug

As the title says, the ParquetRecordBatchReader can not recognize duration type written by pandas or polars.

To Reproduce

First, we should prepare parquet file

import polars as pl
from datetime import timedelta

df = pl.DataFrame({
    "a":  [timedelta(days=1) for _ in range(100)]
})

df.write_parquet("./test.parquet")

Then, read in rust arrow-rs:

fn main() -> Result<()> {
    // Create parquet file that will be read.

    let path = "./test.parquet";
    let file = File::open(path).unwrap();

    let parquet_reader = ParquetRecordBatchReaderBuilder::try_new(file)?
        .with_batch_size(8192)
        .build()?;

    let mut batches = Vec::new();

    for batch in parquet_reader {
        batches.push(batch?);
    }

    println!("{:#?}", batches[0].schema());

    Ok(())
}

finally we get:

Schema {
    fields: [
        Field {
            name: "a",
            data_type: Int64,
            nullable: true,
            dict_id: 0,
            dict_is_ordered: false,
            metadata: {},
        },
    ],
    metadata: {},
}

Expected behavior

polars result:

shape: (100, 1)
┌──────────────┐
│ a            │
│ ---          │
│ duration[μs] │
╞══════════════╡
│ 1d           │
│ 1d           │
│ 1d           │
│ 1d           │
│ 1d           │
│ …            │
│ 1d           │
│ 1d           │
│ 1d           │
│ 1d           │
│ 1d           │
└──────────────┘

pandas result:

        a
0  1 days
1  1 days
2  1 days
3  1 days
4  1 days
..    ...
95 1 days
96 1 days
97 1 days
98 1 days
99 1 days

[100 rows x 1 columns]

Additional context