[ENH] Return to float16 for streamline data

[36000]

Not linearizing because it only decreased disk size by around 30% for 0.1mm tolerance

[Copilot]

Pull request overview

This PR aims to reduce streamline storage size by returning to float16 for streamline positions (especially for TRX outputs) and adds safeguards/normalization to keep downstream DIPY operations working with float32.

Changes:

• Updates cuslines compression settings in probabilistic/PTT tracking to emit float16 positions.
• Adjusts TRX dtype handling in tractography/segmentation tasks (positions/offsets).
• Forces resampling inputs to float32 via Streamlines normalization before calling DIPY resampling utilities.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
setup.cfg	Updates gpu extra cuslines pin (currently creates a resolver conflict with core pin).
AFQ/tractography/tractography.py	Sets cuslines tracker compression parameters to use float16 positions.
AFQ/tasks/tractography.py	Changes TRX dtype_dict (offsets dtype) for TRX writing from lazy tractograms.
AFQ/tasks/segmentation.py	Switches TRX streamline positions to float16 and adjusts per-bundle export behavior.
AFQ/recognition/utils.py	Normalizes streamlines to float32 before resampling via DIPY.

---

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