Non-ASCII character in application entity name is disruptive

[Enet4]

This issue already accounts for the fix in #566.
Given the following code:

let mut table = FileMetaTable {
    information_group_length: 0,
    information_version: [0u8, 1u8],
    media_storage_sop_class_uid: "1.2.840.10008.5.1.4.1.1.7".to_owned(),
    media_storage_sop_instance_uid: "2.25.137731752600317795446120660167595746868".to_owned(),
    transfer_syntax: "1.2.840.10008.1.2.4.91".to_owned(),
    implementation_class_uid: "2.25.305828488182831875890203105390285383139".to_owned(),
    implementation_version_name: Some("MYTOOL100".to_owned()),
    source_application_entity_title: Some("RUSTYÀ".to_owned()), // <-- see `À` in there
    receiving_application_entity_title: None,
    sending_application_entity_title: None,
    private_information_creator_uid: None,
    private_information: None,
};

table.update_information_group_length();

let mut buf = vec![b'D', b'I', b'C', b'M'];
table.write(&mut buf).unwrap();

let table2 = FileMetaTable::from_reader(&mut buf.as_slice())
    .expect("Should not fail to read the table from the written data");

assert_eq!(table.information_group_length, table2.information_group_length);

It panics with the following error:

Should not fail to read the table from the written data: DecodeElement {
  source: ReadHeaderTag {
    backtrace: None,
    source: Error {
      kind: UnexpectedEof,
      message: "failed to fill whole buffer"
    }
  }
}

If I add one more character to the source application entity title:

let mut table = FileMetaTable {
    // ...
    source_application_entity_title: Some("RUSTYÀ-".to_owned()), // <--
    // ...
};

It parses correctly but then the information group length no longer matches the version previously calculated.

assertion `left == right` failed
  left: 216
 right: 218

It does not happen if the non-ASCII character is removed. This could have to do with the text encoding/decoding process from dicom-encoding.