Bypass tissue mask creation by requesting brain mask

[paola-g]

Since the WM and CSF signal are precomputed by fmriprep, we could in theory bypass the makeTissueMasks step and just use a user-specified brainmask to create the allMask.

Need to check that there are no other steps depending on the WM and CSF mask. However, some steps use the GM mask, but this is pipeline-dependent.

[paola-g]

The func folder contains a brainmask, that I could use directly instead of asking for a user specified file. For the GM, I could either use the output of freesurfer (in sourcedata/) but it needs conversion, or I can use the probseg GM mask in T1w space and register it to the native bold space (the T1w and the MNI space cases are already covered).

[paola-g]

In the bob branch, I implemented a version that retrieved the brain mask from the func folder unless it is specified by the user. Similarly, for tissue masks, I try to retrieve them from the anat folder unless they are provided explicitly by the user. I still need to test that the resampling did not introduce any errors.

The old behaviour for mask creation can be still accessed with the --legacy flag.