Add surface reconstruction

[chrisgorgo]

• Use FreeSurfer 6 beta
• Run FS by default but allow turning it off on the command line
• Make sure to use all available cores for FreeSurfer
• Use -highres if the input T1 resolution is <1mm
• Use WM and ventricle parcels produced by FS in aCompCor
• Feed multiple T1s to recon-all if available (but only from one session and only if they have the same dimensions/resolution - for longitudinal datasets each session should be analyzed separately)
• Feed T2s to recon-all if available (but only if they have resolution along the coarsest dimension < 1.2mm)
• Replace internal Freesurfer skullstripping with something better (for example ANTs) - see https://github.com/nipy/nipype/blob/master/nipype/workflows/smri/freesurfer/recon.py#L12
• If T2s are available use sqrt(T1*T2) bias field correction method (see HCP pipelines for details)
• Use FreeSurfer implementation of BBR for EPI to T1 coregistration
• Output:
  • SUBJECTS_DIR (with all subjects)
  • fMRI timeseries in fsaverage space

Implement and use reportlets outlined in nipreps/niworkflows#102 and nipreps/niworkflows#103

[oesteban]

Duplicate of #63. But here we have a clear description. I close the old one.