[ENH] Use PyMARE for image-based meta-analyses

[tsalo]

Closes #211, closes #108.

Changes proposed in this pull request:

• Remove ESMA module.
• Use PyMARE for IBMA methods (except RFX GLM, for now).
• Add a new NiMARE test dataset made from the 21-pain studies NiMARE dataset. Instead of full images, however, we have 10x10x10 maps. Also includes transformed maps like z and varcope.
• Add mask and atlas images for new test dataset.

[tsalo]

So far, the main blocker on this has been neurostuff/PyMARE#42, but a simple hack is to just mask out any voxels where any of the input maps has a value of exactly 0. I'm going to push that for now, even though it's a suboptimal solution.

[tsalo]

The IBMA tests are passing locally!!!! 🎉 But we are going to either need a fixed release for PyMARE or pin the dependency to GitHub.

Also the outputs only grab FE stats, so if we really want the RE stats, that will need to be updated. I also haven't done anything as far as permutation tests of MCC of any kind. The standard NiMARE correctors (e.g., FDRCorrector) should work fine, but it's possible that we'll want some native correction methods implemented for the IBMA estimators (like how we have the montecarlo method for most of the CBMA estimators.

[codecov]

Codecov Report

Merging #273 into master will increase coverage by 1.18%.
The diff coverage is 95.93%.

@@            Coverage Diff             @@
##           master     #273      +/-   ##
==========================================
+ Coverage   72.97%   74.15%   +1.18%     
==========================================
  Files          40       39       -1     
  Lines        3882     3711     -171     
==========================================
- Hits         2833     2752      -81     
+ Misses       1049      959      -90     

Impacted Files	Coverage Δ
nimare/workflows/conperm.py	88.88% <66.66%> (-7.27%)	⬇️
nimare/meta/ibma.py	97.79% <97.27%> (+65.99%)	⬆️
nimare/base.py	66.85% <100.00%> (+0.76%)	⬆️
nimare/cli.py	97.64% <100.00%> (+0.02%)	⬆️
nimare/meta/__init__.py	100.00% <100.00%> (ø)
nimare/extract/utils.py	26.92% <0.00%> (-16.93%)	⬇️
nimare/extract/extract.py	16.84% <0.00%> (-12.11%)	⬇️
nimare/dataset.py	88.15% <0.00%> (-0.88%)	⬇️
nimare/transforms.py	70.40% <0.00%> (-0.45%)	⬇️
nimare/utils.py	93.95% <0.00%> (+0.54%)	⬆️

---

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 bf7f21d...d09ebae. Read the comment docs.