FIX: Minor refactor of DWI/brainmasks utilities and dtypes

[oesteban]

This PR:

• [x] Handles the data dtypes more carefully, avoiding casting to floats whenever possible.
• [x] Instead of normalizing to an arbitrary value (1000), the b=0 volumes are normalized to the first one (which is expected to have the highest median signal).
• [x] Adds a signal_drift output to the RescaleB0 interface to reuse down the line.

Future lines

We should think about dropping FSL MCFLIRT, which is used for HMC of the b = 0 volumes.

Resolves: #90

[pull-assistant[bot]]

Best reviewed: commit by commit

---

Optimal code review plan

     FIX: Minor refactor of DWI/brainmasks utilities and dtypes

Powered by Pull Assistant. Last update a66d685 ... a66d685. Read the comment docs.

[codecov[bot]]

Codecov Report

Merging #101 into master will increase coverage by 0.24%. The diff coverage is 15.00%.

@@            Coverage Diff             @@
##           master     #101      +/-   ##
==========================================
+ Coverage   50.81%   51.06%   +0.24%     
==========================================
  Files          21       21              
  Lines        1222     1218       -4     
  Branches      162      161       -1     
==========================================
+ Hits          621      622       +1     
+ Misses        586      581       -5     
  Partials       15       15              

Impacted Files	Coverage Δ
dmriprep/utils/images.py	15.00% <11.11%> (+1.66%)	:arrow_up:
dmriprep/interfaces/images.py	71.42% <50.00%> (+0.84%)	:arrow_up:

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update a27fc16...a66d685. Read the comment docs.

[oesteban]

Seems to resolve the issue -> https://314-233429127-gh.circle-artifacts.com/0/tmp/THP002/derivatives/dmriprep/sub-THP0005.html#acquisition-GD72_datatype-dwi_desc-brain_run-1_session-THP0005JHU1_suffix-mask