Remove 3.5 support, add tests on other Python versions, and break StanMetaRegression for Python 3.7+

[tsalo]

Closes #63.

Changes proposed:

• Remove the CircleCI config file.
• Add a GHA workflow to run tests on Python versions 3.6-3.9.
  • I adapted nilearn's workflow.

[codecov-commenter]

Codecov Report

Merging #66 (bcee35c) into master (ecc71fa) will decrease coverage by 0.24%.
The diff coverage is 83.15%.

@@            Coverage Diff             @@
##           master      #66      +/-   ##
==========================================
- Coverage   88.87%   88.63%   -0.25%     
==========================================
  Files          10       10              
  Lines         764      774      +10     
==========================================
+ Hits          679      686       +7     
- Misses         85       88       +3     

Impacted Files	Coverage Δ
pymare/results.py	81.81% <73.33%> (+0.37%)	⬆️
pymare/core.py	90.90% <75.00%> (ø)
pymare/estimators/estimators.py	83.65% <78.18%> (-1.00%)	⬇️
pymare/estimators/combination.py	92.00% <87.50%> (+0.16%)	⬆️
pymare/effectsize/base.py	95.65% <92.10%> (ø)
pymare/__init__.py	100.00% <100.00%> (ø)
pymare/effectsize/__init__.py	100.00% <100.00%> (ø)
pymare/effectsize/expressions.py	100.00% <100.00%> (ø)
pymare/estimators/__init__.py	100.00% <100.00%> (ø)
pymare/stats.py	92.30% <100.00%> (ø)
... and 4 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 a794483...bcee35c. Read the comment docs.