Fix bad phase connection when using --denoise-after-combining

[tsalo]

Closes #775. In #679, we added support for complex-valued denoising, but only if --denoise-after-combining isn't used. This removes a connection from when I tried supporting complex-valued denoising when --denoise-after-combining is used.

It looks like this bug was found by running --denoise-after-combining. @mattcieslak any chance we could modify one of the tests to use --denoise-after-combining? Not sure if there are any test datasets with multiple runs.

Changes proposed in this pull request

• Remove expected out_dwi_phase output from merge_dwis node.

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 29.77%. Comparing base (65791db) to head (9d0778c).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #781   +/-   ##
=======================================
  Coverage   29.77%   29.77%           
=======================================
  Files          97       97           
  Lines       14583    14583           
  Branches     1886     1886           
=======================================
  Hits         4342     4342           
  Misses      10115    10115           
  Partials      126      126           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[mattcieslak]

I have always thought that --denoise-after-combining is a bad idea. It was originally included for pipeline testing as kind of a strawman. Maybe the best option would be to just remove it?

[mattcieslak]

also I don't think any of the test datasets have multiple runs :(

[tsalo]

Unfortunately, I don't know enough about it to have an opinion about --denoise-after-combining. It looks like drbuddi_rpe_series has an AP and a PA run of the DWI. Would the parameter be applicable there?