[mt] Cover all regression objectives.

[trivialfis]

• Support multi-target for the following objectives:
  • hinge
  • poisson
  • tweedie
  • pseudohubererror
  • squaredlogerror
• Add tests.

ref #9043

[Copilot]

Pull request overview

Extends XGBoost’s multi-target (multi-output tree) coverage for additional built-in objectives, and adds CPU/GPU Python tests to guard against regressions.

Changes:

• Update regression objectives to correctly handle multi-target gradient shapes/indices (notably Poisson/Tweedie; fix indexing for squaredlog/pseudohuber).
• Add reusable label validation types for Poisson/Tweedie label constraints.
• Add CPU + GPU Python test cases covering the newly supported objectives (and broader regression objective coverage).

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
tests/python/test_multi_target.py	Adds CPU test entry points for additional objectives under multi-target training.
tests/python-gpu/test_gpu_multi_target.py	Adds GPU test entry points for the same objectives under multi-target training.
src/objective/regression_obj.cu	Implements/adjusts multi-target gradient computation for Poisson/Tweedie and fixes per-target gradient indexing for some objectives.
src/objective/regression_loss.h	Introduces Poisson/Tweedie label validation helpers used by objectives.
src/objective/aft_obj.cu	Adds a runtime check clarifying AFT label usage requirements.
python-package/xgboost/testing/multi_target.py	Adds a common regression-objective test helper and new per-objective multi-target test runners.

---

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

[trivialfis]

cc @rongou