Optimize Newton replicated startup paths

[kellyguo11]

Description

Optimize replicated-scene startup for Newton camera and non-camera tasks while preserving heterogeneous-environment support.

What changed

• Construct Shadow Hand reorientation assets and cameras inside ReplicateSession, so backend replication sees the complete asset queue.
• Avoid concrete USD clones for pure headless Newton and Newton Warp runs while retaining them for PhysX, RTX, GUI/offscreen output, visualizers, and explicit USD clone consumers.
• Resolve camera prototypes through the clone plan, cache shared camera intrinsics, and avoid redundant stage traversal.
• Import clone-external Newton USD roots directly instead of traversing the replicated environment tree from the pseudo-root.
• Scope collision-approximation, mesh, transform, color, and heightfield work to imported roots and selected solver shapes.
• Cache articulation target-mode resolution and reuse joint-wrench and site-frame view work.
• Resolve semantic labels through the clone plan, reading labels once per prototype while preserving per-environment instance identities.
• Preserve heterogeneous tasks by retaining per-source builders, source selection, collision compatibility checks, and safe fallbacks.

Dependencies: none.

Performance

Measured against upstream/develop at e3c7dd87b6 on an RTX PRO 6000 Blackwell with 4096 environments and warm asset/kernel caches. These are standardized isaaclab benchmark startup cProfile totals from launch phases through the completed first step.

Task	Presets	Before	After	Change
Shadow Hand vision	newton_mjwarp,newton_renderer	38.524 s	16.152 s	-22.372 s (-58.1%)
Kuka-Allegro lift camera	newton_mjwarp,newton_renderer,single_camera,rgb64	29.385 s	27.831 s	-1.554 s (-5.3%)
G1 rough terrain, no camera	newton_mjwarp	20.118 s	12.912 s	-7.206 s (-35.8%)

Environment-creation-only timing improved from 35.832 s to 13.574 s for Shadow (-62.1%), 25.097 s to 23.503 s for Kuka (-6.4%), and 18.294 s to 11.087 s for G1 (-39.4%). First-step time remained effectively unchanged for all three tasks.

Kuka improves less because its heterogeneous multi-object scene still requires per-source builders, compatibility checks, collision fallback handling, and four contact-sensor constructions. The homogeneous fast path cannot collapse those task-specific costs.

The standardized profiled totals should not be compared directly with earlier unprofiled prototype measurements such as 9.898 s because cProfile instruments Python-heavy scene traversal and the measurement boundaries differ.

Type of change

• Bug fix (non-breaking change which fixes an issue)
• Performance improvement

Validation

• 33 focused tests passed previously on this exact commit, including Kuka heterogeneous RGB and semantic-segmentation golden tests.
• Repeated the focused cloner, collision, articulation, and segmentation suite: 31 passed.
• Verified the new regression tests fail against unmodified upstream/develop.
• Isaac-Lift-KukaAllegro-Camera: RSL-RL, 16 environments, Newton MJWarp plus Newton renderer, two learning iterations completed.
• Isaac-Reorient-Cube-Shadow-Camera-Direct: RSL-RL, 16 environments, Newton MJWarp plus Newton renderer, two learning iterations completed.
• All three 4096-environment benchmarks completed scene creation and the first step.
• uv run --frozen python tools/changelog/cli.py check develop: passed.
• uv run --frozen isaaclab -f: passed.
• The legacy source/isaaclab/test/sim/test_cloner.py requires the full Isaac Sim runtime and cannot collect in this kitless worktree. PhysX plus RTX benchmarking remains an explicit CI or reviewer validation request.

Screenshots

Not applicable.

Checklist

• I have read and understood the contribution guidelines
• I have run the pre-commit checks
• Documentation changes are not required for this internal startup path
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• I have added a changelog fragment under source/<pkg>/changelog.d/ for every touched package
• My name already exists in CONTRIBUTORS.md

[kellyguo11]

Closing this duplicate PR; development and review will continue in #6928.