Added code to clip out-of-dwi fiber coordinates

[soichih]

Updated clipping code to handle fiber coordinates overflow (the previous patch only deals with underflow)

[francopestilli]

@soichih how do the results of the model (say weights and rmse), built with this new changes, compare to the results in the published paper? Have you tested that already?

[soichih]

@francopestilli I just ran a few more tests to compare the results and they are very close; within a usual stochastic margin. Once this PR is merged, I am planning to do more thorough testings before making a new release for app-life.

[soichih]

More comparison
1.7

master

29580 ABIDEII

[francopestilli]

@soichih ok let's try this merge we can revert if necessary.

[ccaiafa]

Hi @soichih, as I remember, with STN and HCP we haven't had any underflow or overflow problem. I remember that when I changed ceil to round, the new rmses were a little bit better (smaller).