Handle bytes identifier as UTF-8 strings

Previously, we visited either strings or bytes so that the visitor had
to handle both cases.  With this change, we always visit strings so that
the bytes visitor function can be optimized out. This significantly
reduces binary size.

src/de.rs

@@ -808,9 +808,16 @@ impl<'de, 'a> de::Deserializer<'de> for &'a mut Deserializer<'de> {
     {
         let major = self.peek_major()?;
         match major {
-            MAJOR_STR => self.deserialize_str(visitor),
+            MAJOR_BYTES | MAJOR_STR => {
+                // Rust identifiers are always valid UTF-8 so we can assume that bytes are
+                // UTF-8-encoded strings.  This has the benefit that we only need a mapping from
+                // strings to fields (and the mapping from bytes to fields can be optimized out).
+                let length = self.raw_deserialize_u32(major)? as usize;
+                let bytes: &'de [u8] = self.try_take_n(length)?;
+                let string_slice = core::str::from_utf8(bytes).map_err(|_| Error::DeserializeBadUtf8)?;
+                visitor.visit_borrowed_str(string_slice)
+            }
             MAJOR_POSINT => self.deserialize_u64(visitor),
-            MAJOR_BYTES => self.deserialize_bytes(visitor),
             _ => Err(Error::DeserializeBadMajor),
         }
     }