[isaac] Post-Phase-2 cleanup: drop PR/review meta from test module docstrings

[yinsong1986]

Context

Tracker for cleanup ask from cagataycali on PR #46 (procedural USD builders):

After we merge and verify isaac sim, we should remove comments from the code as cleanup, not blocking
— #46 (comment)

The [isaac] Phase-1 test files under strands_robots_sim/isaac/tests/ carry module-level docstrings that explain why each pin exists by referencing the originating PR # and reviewer comment, e.g.:

• test_procedural_kinematic_guard.py — opens with "cagataycali's review on PR feat(isaac): procedural USD builders (SO-100, Panda, G1) (3/5 of #31 split) #46 (procedural USD builders) flagged that…"
• test_procedural_g1_dof.py — "R2 review on feat(isaac): IsaacSimulation backend skeleton + entry-point registration (R7) #31 surfaced that procedural.py's G1 docstring / inline comment claimed '29 DOF'…"
• test_entrypoint.py — "This is the PR-1 slice of feat(isaac): IsaacSimulation backend skeleton + entry-point registration (R7) #31 (see issue [isaac] R7 Phase 1: umbrella tracker for split of #31 #42)…"

These references were useful as the split PRs (#44 / #45 / #46 / #47 / #48) landed — they document the rationale each pin was reacting to. After the chain merges and Isaac Sim is verified end-to-end (Phase 2 / #50 / PR #51), the PR-meta context becomes redundant: the tests themselves stand on their own as behavioural pins, and git blame plus the merge commits preserve the audit trail.

Scope

Once both of these are true, do a single cleanup pass on strands_robots_sim/isaac/tests/:

1. The Isaac PR chain (feat(isaac): [isaac] extras + entry-point + PEP 562 lazy import (1/5 of #31 split) #44 → docs(isaac): backend reference + Phase 1 status banner (5/5 of #31 split) #48) has fully merged into main.
2. Phase 2 (#50 / PR feat(isaac): URDF / MJCF / USD loaders → ProceduralRobot (closes #50) #51 — URDF/MJCF/USD loaders) has been merged and an end-to-end Isaac Sim instantiation run has been verified (the kinematic guard's documented G1 defect either fixed or reproducibly observable through the loader path, etc.).

Proposed cleanup

Per file, rewrite the module docstring to a content-only summary — describe the pin (what behaviour it locks in, why a regression would be a problem) without the PR / reviewer / split-slot references. Inline references to specific PRs / issues that are still load-bearing for future reference (e.g. "see #50 for Phase 2 plan") are fine to keep.

Concretely:

• test_procedural_kinematic_guard.py — drop the cagataycali's review on PR #46 lead paragraph; keep the three pinned-property bullets.
• test_procedural_g1_dof.py — drop the R2 review on #31 surfaced… paragraph; keep the actual claim ("21-DOF; pinned so the comment / docstring don't drift").
• test_entrypoint.py — drop the PR-1 slice of #31 / see issue #42 paragraph; keep the contract description and the deferred-coverage note.

Same treatment for any inline # review #N said… comments in non-test files under strands_robots_sim/isaac/ if they survive that far.

Non-goals

• Don't change test behaviour or assertions.
• Don't drop pins. The defect-trigger expectations (e.g. duplicate parent->child body edges, 21 DOF, unitree_g1 name) are the value of these tests.
• Don't backfill / rewrite history; the cleanup is forward-only.

Out-of-scope deferral note

Filed in response to a non-blocking review nit on #46 — the comment explicitly said "not blocking", so #46 ships as-is and this issue tracks the post-Phase-2 follow-up.

[yinsong1986]

Status check — gating conditions not yet met

The issue scope explicitly conditions this cleanup on both of:

1. The Isaac PR chain (feat(isaac): [isaac] extras + entry-point + PEP 562 lazy import (1/5 of #31 split) #44 → docs(isaac): backend reference + Phase 1 status banner (5/5 of #31 split) #48) fully merged into main.
2. Phase 2 ([isaac] R7 Phase 2 follow-up: load ProceduralRobot from URDF / MJCF / USD instead of hardcoded builders #50 / PR feat(isaac): URDF / MJCF / USD loaders → ProceduralRobot (closes #50) #51) merged + an end-to-end Isaac Sim run verified.

As of today (state on upstream/main at 0341304):

PR / Issue	State
#44 — [isaac] extras + entry-point + lazy import (1/5)	✅ Merged
#45 — IsaacConfig dataclass (2/5)	✅ Merged
#46 — procedural USD builders (3/5)	⏳ Open
#47 — IsaacSimulation backend (4/5)	⏳ Open
#48 — backend reference + Phase 1 banner (5/5)	⏳ Open
#50 / #51 — Phase 2 URDF/MJCF/USD loaders	⏳ Open (#51)

Two of the three target files (test_procedural_kinematic_guard.py, test_procedural_g1_dof.py) don't exist on main yet — they live on the pr3/isaac-procedural branch and ride in with #46. Only test_entrypoint.py is on main, and its docstring still carries the PR-1 slice of #31 (see issue #42) paragraph the issue calls out.

Doing a partial cleanup on test_entrypoint.py alone now would:

• Defeat the "single pass" intent the issue describes.
• Touch a file the still-open chain may need to retouch (lower-risk than rebase, but still adds churn for pr3/.../pr4/.../pr5/... reviewers tracking inline-thread anchors).
• Trigger a second cleanup pass once feat(isaac): procedural USD builders (SO-100, Panda, G1) (3/5 of #31 split) #46–docs(isaac): backend reference + Phase 1 status banner (5/5 of #31 split) #48 + feat(isaac): URDF / MJCF / USD loaders → ProceduralRobot (closes #50) #51 land, which is exactly what the issue is filed to avoid.

Recommendation

Hold this issue until after:

• feat(isaac): procedural USD builders (SO-100, Panda, G1) (3/5 of #31 split) #46, feat(isaac): IsaacSimulation backend (lifecycle + observation + render) (4/5 of #31 split) #47, docs(isaac): backend reference + Phase 1 status banner (5/5 of #31 split) #48 are merged and the Phase-1 surface is in main.
• feat(isaac): URDF / MJCF / USD loaders → ProceduralRobot (closes #50) #51 is merged and an end-to-end Isaac Sim instantiation run has been verified (per the issue: kinematic guard's documented G1 defect either fixed or reproducibly observable through the loader path).

I can pick this up immediately once those land. No changes shipped on this run; the issue stays open and pinned to its post-Phase-2 trigger.