Do not inherit IBMAEstimator's aggressive_mask from previous Datasets

[tsalo]

Closes #651.

Changes proposed in this pull request:

• Remove the aggressive_mask key at the beginning of MetaEstimator._preprocess_inputs() if it's defined. Basically, this redefines aggressive_mask from scratch any time fit is called.
• Mask data with the aggressive mask only if that aggressive mask is defined.
  • At first glance, this should be a bug that's come up already, and yet it hasn't. I assume there's a reason why CBMAEstimators loading ma_maps haven't failed due to a missing aggressive_mask key in inputs_, but I don't know what it is at the moment.

[codecov]

Codecov Report

Merging #652 (86b5878) into main (8f166ab) will increase coverage by 0.01%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main     #652      +/-   ##
==========================================
+ Coverage   84.88%   84.89%   +0.01%     
==========================================
  Files          40       40              
  Lines        4386     4389       +3     
==========================================
+ Hits         3723     3726       +3     
  Misses        663      663              

Impacted Files	Coverage Δ
nimare/base.py	88.54% <100.00%> (+0.15%)	⬆️

---

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 8f166ab...86b5878. Read the comment docs.