Reject empty TXT records in Txt::parse (#709)

RFC 1035 section 3.3.14 requires "one or more" character strings, so
rdlength=0 is invalid. `check_slice` (used by `from_octets` /
`from_slice`) already enforced this; `parse` did not, allowing
construction of a `Txt` that panics in `as_flat_slice`.

Replace the manual `CharStr::skip` loop with a single `check_slice`
call so all construction paths share the same validation.

The zero-length unknown-rdata test now uses TYPE65534 instead of TXT
since zero-length TXT is correctly invalid.

src/rdata/rfc1035/txt.rs

@@ -185,12 +185,9 @@ impl<Octs> Txt<Octs> {
         Octs: AsRef<[u8]>,
     {
         let len = parser.remaining();
-        LongRecordData::check_len(len)?;
         let text = parser.parse_octets(len)?;
-        let mut tmp = Parser::from_ref(text.as_ref());
-        while tmp.remaining() != 0 {
-            CharStr::skip(&mut tmp)?
-        }
+        Txt::check_slice(text.as_ref())
+            .map_err(|e| ParseError::Form(e.into()))?;
         Ok(Txt(text))
     }
 
@@ -1147,6 +1144,12 @@ mod test {
         );
     }
 
+    #[test]
+    fn txt_empty_rejected() {
+        let mut parser = Parser::from_static(b"");
+        assert!(Txt::<&[u8]>::parse(&mut parser).is_err());
+    }
+
     #[test]
     fn txt_display() {
         fn cmp(input: &[u8], output: &str) {

test-data/zonefiles/unknown-zero-length.yaml

@@ -1,10 +1,10 @@
 origin: example.com.
 zonefile: |
-  example.com. 3600 IN TXT \# 0
+  example.com. 3600 IN TYPE65534 \# 0
 result:
   - owner: example.com.
     class: IN
     ttl: 3600
     data: !Unknown
-      rtype: Txt
+      rtype: TYPE65534
       data: