[FIX] More reliable brain masking/FA

[36000]

(1) Currently, with DIPY defaults, DKI does not fit in many voxels in HCP subjects. I am looking into fixing this, but for now, let's just use DTI FA
(2) Adds MSDKI metrics to defaults. They calculate easily and provide some complementary information.
(3) Changes default brain mask from brainchop to synthseg. I have just noticed synthseg is more reliable.

[Copilot]

Pull request overview

This PR updates pyAFQ’s default structural masking and diffusion scalar outputs to improve robustness in common datasets (e.g., HCP), while also simplifying some model-fitting internals.

Changes:

• Switch default T1w brain masking to derive a binary mask from SynthSeg2 segmentations (instead of Brainchop/Mindgrab).
• Update default scalar list for multi-shell data to use DTI FA/MD and add MSDKI metrics (MSD/MSK), avoiding DKI FA by default.
• Remove unused/duplicative code paths (e.g., unused MS-MT intermediate, private DKI _fit helper).

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 4 comments.

File	Description
AFQ/tasks/structural.py	Uses SynthSeg2 output to construct the default T1w brain mask; updates SynthSeg task input to avoid masked-T1 dependency.
AFQ/tasks/data.py	Adjusts DKI fitting call site, adds MSDKI metrics to defaults, and adds thresholding to MSDKI outputs.
AFQ/models/msmt.py	Removes an unused A_outer computation.
AFQ/models/dki.py	Removes a private _fit helper no longer used by the tasks layer.

---

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

[36000]

@arokem this is ready for review/merge, then we can make a 3.2 release!