[FIX/ENH] VOF HBN changes

[36000]

Includes these changes:

• max numba thread count is explicitly bounded, avoiding an error where you give numba too many threads
• Gaussian weighting has an extra check and is switched to float64. This changes default behavior, which was to use float32. Float32 only ran into problems with large numbers of streamlines (and thus small weights). So, for normal numbers of streamlines, this should only change default behavior slightly.
• t1w/log(bo) is added as an option, but not made default
• Fixes a bug in our brainchop models, which I believe was in here since 3.0. Luckily brainchop has not been our default.
• Adds an option to use all 3 NNs to make a wm mask

[Copilot]

Pull request overview

This PR adds a new derived scalar in AFQ.tasks.data to compute a T1w-over-log(b0) image intended as an R1 proxy, and wires it into the default data task plan so it can be produced as part of standard runs.

Changes:

• Added a new t1w_over_log_b0 calculation task that computes T1w / log(mean b0) after resampling T1w into b0 space.
• Registered the new task in get_data_plan so it runs alongside existing scalar generation (e.g., t1w_over_b0).

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 7 out of 7 changed files in this pull request and generated 3 comments.

Comments suppressed due to low confidence (1)

AFQ/tasks/tissue.py:126

• The ValueError message is now outdated: it doesn’t mention the newly supported option 'multiaxial+brainchop+synthseg', which can confuse users when they pass an invalid value.

    raise ValueError(
        "pve must be a PVEImage, PVEImages, 'synthseg', or 'multiaxial+brainchop'"
    )

[36000]

@arokem this is ready for review / merge

[arokem]

Any chance to add a test of this?

[36000]

Yes, I can do that