Improve incorrect type handling on columnar

[PSeitz]

The way incorrect types on typed columns are handled in columnar is a footgun. We should ignore it or return an error.

#[test]
fn test_date_range_query() {
    let mut schema_builder = Schema::builder();
    let date_field = schema_builder.add_date_field("date", FAST);
    let schema = schema_builder.build();

    let index = Index::create_in_ram(schema.clone());
    
      // This is adds a string and creates a string column
      index_writer
          .add_document(doc!(date_field => "2022-12-01T00:00:01Z"))
          .unwrap();
          
      let mut index_writer = index.writer_with_num_threads(1, 50_000_000).unwrap();
      // Parses the string to a Date. Now we have 2 columns under "date"
      let doc = TantivyDocument::parse_json(&schema, r#"{ "date": "2022-12-01T00:00:01Z" }"#)
          .unwrap();
      index_writer.add_document(doc).unwrap();
      
      // Adds a Date
      index_writer
          .add_document(doc!(date_field => DateTime::from_utc(
              OffsetDateTime::parse("2023-12-01T00:00:01Z", &Rfc3339).unwrap(),
          )))
          .unwrap();

      index_writer.commit().unwrap();
    
}