Non-blocking read will return the amount of data available without
blocking to wait for the full requested size.

The downside is that now it souphttpsrc needs to have a waiting
mechanism in case there is no data available yet to avoid busy
looping arond the inputstream.