honour response charset in StringDecoder

[alhudz]

Repro: a String-returning method against a server that replies Content-Type: text/plain; charset=ISO-8859-1 with body byte 0xE9.
Expected: é.
Actual: U+FFFD, because the byte is run through a UTF-8 reader.
Cause: StringDecoder.decode reads the body with asReader(Util.UTF_8), so the charset response.charset() already resolves from Content-Type is dropped. DefaultDecoder extends it, so this is the default behaviour for every non-UTF-8 String response.
Fix: decode with response.charset(), which already falls back to UTF-8 when none is declared, matching the JSON decoders (jackson, json, fastjson2, jackson-jr).

[alhudz]

any update?

[velo]

Approved via agent-assisted review: change includes tests, no backwards-compatibility break, no security concerns, CI green.