[Bug] RpcClientUtils.encodeBody overwrites the caller's ByteBuffer mark

Description

RpcClientUtils.encodeBody in the remoting module mutates buffer state it does not own:

} else if (body instanceof ByteBuffer) {
    ByteBuffer buffer = (ByteBuffer)body;
    buffer.mark();
    byte[] data = new byte[buffer.remaining()];
    buffer.get(data);
    buffer.reset();
    return data;
}

The method restores the buffer's position (via mark/reset) but overwrites whatever mark the caller had previously set. A caller that did buffer.mark() at position 1, advanced to position 3, then passed the buffer through encodeBody gets back a buffer whose mark is now 3; a later buffer.reset() jumps to 3 instead of 1. Encoding a body therefore has an observable side effect on the caller's buffer beyond reading it.

Additionally, when the caller's position/mark/limit are what they are, the encode itself must only read the remaining bytes — which duplicate() achieves without touching the original buffer's position, limit or mark at all.

Expected behavior

encodeBody should be side-effect free on the passed-in ByteBuffer: read the remaining bytes from buffer.duplicate() and leave the original buffer's position, limit and mark untouched.