Ensure all object features get exported to table

[k-dominik]

Everytime I look at exportFile.py I want to rewrite it. In lieu of time, another hack...

Previously, only features computed in the 1st timeframe would get a column in the export table. One would expect that with this bug there would be none ;). Global features are always computed -> produce an (empty) table in every frame, but local ones, only if there are objects.

This is more a bandaid then a proper fix. It would probably be better to go from selected features (from the dialog) for
column names, but currently the plugin information is stripped from those :(

CC: @oanegros (thank you for pointing it out and helping to debug)

[codecov]

Codecov Report

Attention: 3 lines in your changes are missing coverage. Please review.

Comparison is base (e813293) 55.75% compared to head (84dd025) 55.77%.
Report is 12 commits behind head on main.

Files	Patch %	Lines
ilastik/utility/exportFile.py	78.57%	1 Missing and 2 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2814      +/-   ##
==========================================
+ Coverage   55.75%   55.77%   +0.01%     
==========================================
  Files         530      530              
  Lines       62173    62180       +7     
  Branches     8498     8503       +5     
==========================================
+ Hits        34663    34679      +16     
+ Misses      25764    25756       -8     
+ Partials     1746     1745       -1     

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

[btbest]

Oof, this one's a doozie... Let's say this function is hard to follow, and the naming is all over the place, but I don't feel like the addition in this PR makes it much worse either.

I also couldn't actually try out whether it fixes what it is supposed to fix. Is there a straightforward way to get a dataset where some object features are only computed from time point >0 onwards? I managed to coerce a dataset so that it has no objects in some time points, but that seems to cause errors much earlier in the workflow (see comment below).

Primarily, I would really like computed_features renamed, because it is super confusing to have it existing (and being important to the function) next to computed_feature. I suggest features_by_time. Besides that, basically just typos in the comment...