Unify single-char and multi-char CStrUnit::Char handling.

The two cases are equivalent. C string literals aren't common so there
is no performance need here.

compiler/rustc_ast/src/util/literal.rs

@@ -161,7 +161,6 @@ pub fn from_token_lit(lit: token::Lit) -> Result<LitKind, LitError> {
                         error = Err(LitError::NulInCStr(span));
                     }
                     Ok(CStrUnit::Byte(b)) => buf.push(b),
-                    Ok(CStrUnit::Char(c)) if c.len_utf8() == 1 => buf.push(c as u8),
                     Ok(CStrUnit::Char(c)) => {
                         buf.extend_from_slice(c.encode_utf8(&mut [0; 4]).as_bytes())
                     }