feat(deps): make torch an optional dependency, only required for sensor evaluation

[janickm]

Summary

• Make torch an optional dependency -- users who only need data reading (ncore.data) no longer need PyTorch installed
• ncore.sensors now uses lazy imports via __getattr__, importing torch only when a sensor class is accessed
• Clear ImportError message guides users to install ncore[sensors] if torch is missing
• Wheel updated: torch moved from requires to extra_requires = {"sensors": ["torch"]}
• All data tests pass without torch at import time; all sensor tests pass with torch available

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[janickm]

/ok to test 298f045

[janickm]

/ok to test de720d9

[janickm]

Postponed -- needs more design work around packaging boundaries before making torch optional.