Encode the char after an unpaired high surrogate in writeUtf8(...) - #18252

Motivation:

When `ByteBufUtil.writeUtf8(...)` finds a high surrogate that is not
followed by a low surrogate, it writes `?` and then writes the next char
cast to a `byte`. That only works for ASCII. A non-ASCII char is
truncated to its low byte, and a surrogate pair that follows is replaced
by `??`. `String.getBytes(UTF_8)` writes `?` and then encodes the next
char normally:

```
input                   String.getBytes(UTF_8)   writeUtf8(...)
"a\uD800éb"        61 3f c3a9 62            61 3f e9 62 (invalid UTF-8)
"a\uD800中b"        61 3f e4b8ad 62          61 3f 2d 62
"a\uD800😀b"  61 3f f09f9880 62        61 3f 3f 3f 62
```

`Unpooled.copiedBuffer(CharSequence, UTF_8)` and
`TextWebSocketFrame(String)` go through this code. `QueryStringEncoder`
has a copy of the same logic and percent-encodes only the low byte of
the next char.

Modification:

- Write `?` for the unpaired high surrogate only and encode the next
  char in the next iteration, in all `writeUtf8(...)` variants, in
  `utf8Bytes(...)` and in `QueryStringEncoder`.
- Add tests to `ByteBufUtilTest` and `QueryStringEncoderTest`.

Result:

`writeUtf8(...)`, `utf8Bytes(...)` and `QueryStringEncoder` agree with
`String.getBytes(UTF_8)` for these inputs.

Co-Authored-By: Claude Opus 5.5 <[email protected]>