Release 0.6.0 - #382

Summary

Bumps brainstat.__version__ to 0.6.0 and adds a project CHANGELOG.md documenting the work folded in across #378 / #379 / #380 / #381.

This is a minor release — see the Backwards-incompatible changes section of the changelog for the user-visible behaviour shifts (BigBrain histology profile inversion, SLM input coercion, numpy/pandas pins).

What landed (full detail in CHANGELOG.md)

Statistics-core bug fixes

• yeo7 off-by-1 indexing in SLM._surfstat_to_brainstat_rft (Retrieval of yeo7 information has off-by-1 error in python code #347)
• cmath.sqrt → math.sqrt in SLM.py
• duplicate fetch_template_surface import dropped
• sys.exit in _t_test.py replaced with raise ValueError
• float(...) on size-1 ndarrays replaced with .item() (numpy 2 compat)

Public-API contract fixes (#342)

• SLM.__init__ coerces contrast to numeric; rejects object/categorical with a clear TypeError
• SLM.__init__ coerces integer 0/1 mask to bool
• surf=None + correction='rft' error now points users at correction='fdr'

Behaviour change (#274)

• read_histology_profile now inverts BigBrain intensities by default (invert=True); invert=False recovers prior behaviour. MPC/gradient outputs are unaffected.

New features

• fetch_parcellation(template, "destrieux", ...) returns the FreeSurfer Destrieux 2009 parcellation (no parcellation for GLM with Freesurfer. #343)

Maintenance

• CI fully restored across Python (3.10/3.11 × 3 OS) and MATLAB (R2022a/latest) (Fix CI (Python + MATLAB) and three small bugs in stats #379)
• Hardened MATLAB Neurosynth fetcher with zip integrity check + retry (Neurosynth data failed to get fetched #341)
• MATLAB R2019b/R2020a kept buildable: peak_clus.column_vector no longer uses mustBeVector ([FIX] column_vector uses mustBeVector from MATLAB 2020b+  #285)
• # type: ignore removed from SLM.py and _multiple_comparisons.py; mypy no longer crashes on the class-body imports ([REF] Remove #type: ignore from SLM class #199)

Tests / coverage

• 276 tests pass, 1 platform-skipped on the local sweep.
• 4 new volumetric tests cover the contrast / mask / FDR paths from Volumetric mixed effects #342.
• Total coverage: 62% (a couple of context modules and the data-generator scripts pull the average down; the actively maintained statistics core sits between 64% and 95%).

How to publish

Once this PR merges:

1. Tag v0.6.0 on master to trigger tagged_release.yml (PyPI + MATLAB artifact + GitHub Release).
2. Verify the release notes on GitHub mirror the changelog entry.

Out of scope (still tracked)

• Volumetric cluster-level RFT/TFCE in SLM (the ambitious half of Volumetric mixed effects #342)
• Hosted Desikan-Killiany atlas in fetch_parcellation (companion to no parcellation for GLM with Freesurfer. #343)
• Lifting the setuptools<81 / numpy<2 / pandas<2.1 pins once abagen is patched
• Re-enabling the commented-out mypy CI job once the ~380 latent typing errors are triaged