fix(conformance): keep the default adapter probe envelope-neutral

[doublewhy]

Plain-language summary

• Context: RUN-314 (Define the reference-backend contract and conformance evidence (RUN-314) #197) requires the released conformance harness to be invocable by an independently installed backend without private knowledge; the test(runtime): characterize backend result integrity #1150 falsification matrix noted run_target_conformance with default arguments fails even the honest reference backend.
• Problem: Since authored OS identity started carrying through realization (feat: carry authored OS identity through realization #1141), the default probe scenario's os: linux demands exact SEM-218 os-family corroboration that no hermetic in-process backend declares — so target-provisioning and target-snapshot silently fail for every honest backend (reference and stub alike), and nothing pinned them.
• Fix: Drop the authored OS from the default probe scenario (mirroring the canonical cross-backend-minimal decision that leaving os unset keeps a scenario admissible on more than one envelope) and pin both adapter probes passing for every hermetic backend.

Issue mapping

• Advances Define the reference-backend contract and conformance evidence (RUN-314) #197 (RUN-314) — specifically the "released conformance harness invocable by a downstream package" criterion.
• Follows through on the residue of Target conformance provisioning probe assumes every backend can realize an arbitrary reference scenario (breaks fixed-topology / simulation backends) #663 recorded in test(runtime): characterize backend result integrity #1150's module docstring; Target conformance provisioning probe assumes every backend can realize an arbitrary reference scenario (breaks fixed-topology / simulation backends) #663's reference_scenario= runner parameter remains the escape hatch for bounded/fixed-topology backends.

Summary

• raes_conformance/conformance/target_probes.py: the default conformance scenario no longer authors os: linux; the comment records why (an authored OS is an exact realization requirement, not adapter mechanics — the SEM-218 probes own that concern).
• tests/test_reference_backend_conformance.py: new test_adapter_probes_pass_for_every_hermetic_backend pins target-provisioning and target-snapshot passing for the reference and stub targets, so the probe cannot silently regress again.
• The pinned fail-closed posture is untouched: report.passed stays False for the reference target (opacity and constructive-envelope probes still report unsupported), exactly as test_reference_target_requires_opacity_probe_for_positive_declaration requires.

Compatibility

• No SDL field, schema, portable contract, or CLI changes. The default probe scenario is a harness input, not a published contract; callers who supplied their own reference_scenario are unaffected.

Verification

• Reference and stub targets both pass target-provisioning and target-snapshot; full report posture unchanged.
• nox -s tests: 7,012 passed; coverage gate green.
• Ruff format and lint clean; tools/check_repo_policy.py pass.

🤖 Generated with Claude Code