Fix checking for correct sample length

[hagenw]

In ir_correct_distance() we are checking if conf.N is long enough for the needed delay. Delay is in seconds, but it is directly substracted from conf.N which is in samples.

This pull request should fix that.

[fietew]

It looks good, so far. I would suggest to change the if-condition into ir_origlength+delay*fs>conf.N. It took me a while to figure out the meaning of the current if-condition. This would be also in better alignment with the warning function-call afterwards: there it is also ir_origlength+delay*fs.

[hagenw]

Thanks, I updated the commit accordingly.