[Rust] SchemaBuilder accepts blank or duplicate nested ROW field names

### Search before asking

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

### Fluss version

main (development)

### Please describe the bug 🐞

The Rust `SchemaBuilder` accepts nested `ROW` types whose field names are blank or duplicated, while the Java `RowType` rejects these definitions.

For example, both schemas currently build successfully:

```rust
Schema::builder()
.column(
"payload",
DataTypes::row(vec![DataTypes::field(" ", DataTypes::int())]),
)
.build()
.unwrap();

Schema::builder()
.column(
"payload",
DataTypes::row(vec![
DataTypes::field("value", DataTypes::int()),
DataTypes::field("value", DataTypes::string()),
]),
)
.build()
.unwrap();
```

The problem also applies to `ROW` types nested recursively inside `ARRAY`, `MAP`, or another `ROW`.

This causes Gateway table creation with `validate_only=true` to return 200 for a schema that Fluss Server rejects during actual creation.

Related discussion:
https://github.com/apache/fluss/pull/4107#discussion_r3862660463

### Expected behavior

`SchemaBuilder::build()` should reject blank or duplicate nested `ROW` field names, consistently with the Java implementation.

### Solution

Recursively validate data types during schema construction and reject:

- field names containing only whitespace;
- duplicate field names within the same `ROW`.

Add fluss-rust unit tests and a Gateway `validate_only` regression test.

### Are you willing to submit a PR?

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

Start at the Rust SchemaBuilder and data-type construction path, then compare its nested ROW validation with the Java RowType behavior described in the issue. Add fluss-rust unit coverage and a Gateway validate_only regression test; done means blank or duplicate nested ROW field names are rejected recursively, including inside ARRAY, MAP, and ROW types.

Written by the indexing model from the issue text.