rbnrmf bug in energy binning

I made an error in the new energy binning option - I should be averaging the response rather than summing it so the resulting responses were too large by the binning factor. Also, there was a one-off problem for responses with zero-based channels. Checked in fixes.