Task: Implement Hotfix: Windows build fails on std::getenv deprecation

Replace every std::getenv call in test code that Windows compiles with ores::testing::make_nats_options() – a shared helper in ores.testing that reads the ORES_NATS_* variables through ores::platform::environment::environment::get_value_or_default – so the -Werror deprecation failure on Windows Clang goes away:

• the codegen template ores.cpp.eventing-integration-test.nats_integration_test.org (tangled to cpp_nats_integration_test.cpp.mustache), which feeds ~85 generated *_eventing_integration_tests.cpp files across ores.iam, ores.refdata, ores.reporting, and ores.trading;
• the two hand-written ores.synthetic service tests (feed_config_handler_tests.cpp, folder_feed_control_handler_tests.cpp), which carry a copy of the same env() lambda.

systemd_notify.cpp already uses the ores.platform wrapper; no change.

• No std::getenv remains in any C++ source under projects/ (grep clean).
• Generated test files are produced by the updated template: the template-drift check reports zero diff.
• The changed test targets build and pass on Linux; the fix is portable by construction (the wrapper already guards Windows with _putenv_s).

Root cause: MSVC's STL marks std::getenv deprecated in favour of _dupenv_s, and Windows CI compiles with -Werror, so every file that calls it fails the build (first observed in account_type_eventing_integration_tests.cpp:60). The failing files are codegen-generated from the ores.cpp.eventing-integration-test.nats_integration_test archetype, whose template inlines a per-file test_nats_options() helper (built on std::getenv) into every generated test: 85 copies, plus two more hand-written copies in ores.synthetic. The fix belongs in the template, not in the generated files (CI enforces a zero-diff drift check), and the duplicated helper is consolidated into one shared helper rather than swapped file by file.

Steps:

1. Add ores::testing::make_nats_options(): a header-only helper in ores.testing (nats_options_helper.hpp) that reads the ORES_NATS_* variables via ores::platform::environment::environment::get_value_or_default with the same defaults the per-file helpers used.
2. Edit the org template: drop the inline test_nats_options() block, include the helper header, call ores::testing::make_nats_options() at the client construction site.
3. Re-tangle: compass build --direct tangle_codegen_templates.
4. Regenerate per component with the codegen (--address ores.cpp.eventing-integration-test) and inspect the diff: every generated file should show only the include swap and the call-site change.
5. Replace the two ores.synthetic service tests' local test_nats_options() with the shared helper.
6. Verify: grep for remaining std::getenv, build the affected test targets, run the drift check.

• Review round 1 (2026-08-12): both claude[bot] reviews approve; one non-blocking observation – ORES_NATS_URL set-but-empty loses the localhost fallback with get_value_or_default alone. Fixed in 5995e8d3da: the helper now mirrors the original per-file lambda exactly (read with empty default, then fall back when empty).

Manual QA scenarios (scaffolded via compass add test_scenario, run through the QA Validation Runner panel) that verify this task. Link new ones here as they're created; the scenario doc itself links back via its "Verifies task" field.

The std::getenv copies are gone: one shared ores::testing::make_nats_options() (nats_options_helper.hpp, header-only, reads the ORES_NATS_* variables through the ores.platform environment provider with the original lambda's exact semantics) replaces the 85 inlined helpers in the eventing-integration template output and the 2 hand-written ores.synthetic copies. The template and all 85 regenerated files across iam, refdata, reporting, and trading-cpp changed only in the include swap and call site; the template-drift check passed. Linux build and tests green: ores.iam.core.tests (95s, incl. the 3 regenerated eventing tests) and ores.synthetic.service.tests both pass. Remaining std::getenv in the tree is only inside real_environment_provider.cpp itself. Merged as PR #1981.