Only create Predictors for EFs present in 90% of runs

[adelavega]

Closes #541

I decided on a 90% default

That code in general is not the most efficient and a bit hard to debug, but I can't think of a much better way to do it for now.

At least this will speed things up by only processing useful Features (i.e. those in most runs).

[codecov-io]

Codecov Report

Merging #542 into master will decrease coverage by 2.07%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #542      +/-   ##
==========================================
- Coverage    85.4%   83.32%   -2.08%     
==========================================
  Files          47       38       -9     
  Lines        2281     1649     -632     
==========================================
- Hits         1948     1374     -574     
+ Misses        333      275      -58

Impacted Files	Coverage Δ
neuroscout/populate/extract.py	97.75% <100%> (+0.07%)	⬆️
neuroscout/tests/request_utils.py
neuroscout/tests/api/test_dataset.py
neuroscout/tests/api/test_user.py
neuroscout/tests/api/test_predictor.py
neuroscout/tests/test_views.py
neuroscout/tests/api/test_analyses.py
neuroscout/tests/test_models.py
neuroscout/tests/conftest.py
neuroscout/tests/api/test_run_task.py
... and 1 more

---

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 9b013bf...17842a4. Read the comment docs.

[adelavega]

This seems to be working well