[Bug] [flink] Could not read External Tables

### Search before asking

- [X] I searched in the [issues](https://github.com/apache/incubator-paimon/issues) and found nothing similar.

### Paimon version

0.5

### Compute Engine

Flink 1.16.2

### Minimal reproduce step

[upload_excel.xlsx](https://github.com/apache/incubator-paimon/files/12459344/upload_excel.xlsx)
1. I trans this excel to parquet

df = pd.read_excel(args.file, sheet_name='data')
parquet_file = '/tmp/parquet_file/' + args.name + '.parquet'
df.to_parquet(parquet_file, index=False)`

2. I upload this parquet to "hdfs://draco01:9870/tmp/flink1.17.1/upload_parquet/table_name.parquet"
3. I create external table :

CREATE TABLE default_database.table_name (
col1 string,
col2 bigint,
col3 string,
col4 date,
col5 string,
PRIMARY KEY (col1,col2) NOT ENFORCED
) WITH (
'connector' = 'paimon',
'path' = 'hdfs://draco01:9870/tmp/flink1.17.1/upload_parquet/table_name.parquet'
);

4. Could read the schema:
+------+--------+-------+-----------------+--------+-----------+
| name | type | null | key | extras | watermark |
+------+--------+-------+-----------------+--------+-----------+
| col1 | STRING | FALSE | PRI(col1, col2) | | |
| col2 | BIGINT | FALSE | PRI(col1, col2) | | |
| col3 | STRING | TRUE | | | |
| col4 | DATE | TRUE | | | |
| col5 | STRING | TRUE | | | |
+------+--------+-------+-----------------+--------+-----------+
5 rows in set
5. Could not read the data, it shows below:
`Caused by: org.apache.hadoop.ipc.RpcException: RPC response exceeds maximum data length`

### What doesn't meet your expectations?

Want to know how to fix it.

### Anything else?

_No response_

### Are you willing to submit a PR?

- [ ] I'm willing to submit a PR!

Le rapport fournit upload_excel.xlsx, une conversion de pandas vers Parquet, un chemin HDFS et l’instruction Flink SQL CREATE TABLE comme points d’entrée pour la reproduction. Reproduisez la lecture avec Paimon 0.5 et Flink 1.16.2, puis retracez l’erreur « RPC response exceeds maximum data length » ; le travail est terminé lorsque les données de la table externe peuvent être lues sans cette erreur.

Rédigé par le modèle d'indexation à partir du texte de l'issue.