Unpin aeppl version

[ricardoV94]

Closes #5191

The new version also fixes some test regressions in #5088

[codecov]

Codecov Report

Merging #5203 (30cc813) into main (c22859d) will decrease coverage by 2.98%.
The diff coverage is n/a.

❗ Current head 30cc813 differs from pull request most recent head 4a9f3a4. Consider uploading reports for the commit 4a9f3a4 to get more accurate results

@@            Coverage Diff             @@
##             main    #5203      +/-   ##
==========================================
- Coverage   78.13%   75.14%   -2.99%     
==========================================
  Files          88       88              
  Lines       14181    14174       -7     
==========================================
- Hits        11080    10651     -429     
- Misses       3101     3523     +422     

Impacted Files	Coverage Δ
pymc/tests/sampler_fixtures.py	90.76% <ø> (-0.48%)	⬇️
pymc/variational/updates.py	11.33% <0.00%> (-76.36%)	⬇️
pymc/plots/posteriorplot.py	32.00% <0.00%> (-64.00%)	⬇️
pymc/distributions/transforms.py	44.82% <0.00%> (-47.13%)	⬇️
pymc/data.py	44.44% <0.00%> (-35.12%)	⬇️
pymc/distributions/dist_math.py	59.35% <0.00%> (-28.35%)	⬇️
pymc/tests/conftest.py	66.66% <0.00%> (-23.81%)	⬇️
pymc/distributions/logprob.py	71.90% <0.00%> (-23.15%)	⬇️
pymc/tests/checks.py	80.00% <0.00%> (-20.00%)	⬇️
pymc/backends/ndarray.py	79.64% <0.00%> (-9.74%)	⬇️
... and 4 more

[ricardoV94]

Toggled as a draft while testing xfails that are passing locally but not failing here on Github with the strict-xfail. It was a case of global seeding differing between local testing and when running the whole suite.

[kc611]

Looks like the issue was related to tags and has been fixed already.