This commit fixes the creation of UnknownContentTypeException when it is thrown from HttpMessageConverterExtractor.extractData.
This exception is usually thrown from RestTemplate when HttpMessageConverterExtractor can't extract response body.
In this case HttpMessageConverterExtractor throws UnknownContentTypeException with response body as byte array.
Before this change the returned exception contained incorrect responseBody when InputStream from ClientHttpResponse was not markable (markSupported() is false). This is the usual case for HTTP response's InputStream implementations.
The first byte of response body inside UnknownContentTypeException was missing because it was already read from InputStream.
The responseBody was read from original InputStream, while it should be read from the wrapper of original InputStream.
To a person who will merge: please improve the commit message if possible as I feel difficulties in attempt to correctly and briefly describe the change, that is essentially a bug fix (not a native English speaker).