Previously we use io.LimitReader to limit the number of bytes
from http request. However, there is a subtle difference between
io.LimitReader and io.ReadAll as io.LimitReader will return
a Reader, not a ReadCloser. As such the behavior will actually
be difference in case of error handling (and when to close).
This PR changes io.LimitReader to http.MaxBytesReader
so that the behavior can be preserved (except the number of bytes).
See https://stackoverflow.com/a/52699702
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
c8844a8555