feat: per-workflow tm1_instance result-push target

[nicolasbisurgi]

Summary

Adds a per-workflow tm1_instance setting for the results-push / auto-load target, with a graceful 4-tier resolution chain. Replaces the implicit reliance on default_tm1_instance (now deprecated but still honoured).

Precedence (highest wins):

1. rushti run --tm1-instance <X>
2. Taskfile JSON settings.tm1_instance
3. settings.ini [tm1_integration].tm1_instance
4. settings.ini [tm1_integration].default_tm1_instance (deprecated)

At push time the run logs Result upload target: <instance> (source: <tier>). When all four tiers are empty and push_results is enabled, a reworded warning fires and the run still succeeds.

Soft deprecations (both honoured indefinitely, warn only when actually used):

• settings.ini [tm1_integration].default_tm1_instance → rename to tm1_instance
• rushti tasks push --target-tm1-instance → alias for --tm1-instance

Behavioural change worth highlighting: push_results and auto_load_results set inside a taskfile JSON settings block now actually take effect on the run path. They were silently ignored before because get_effective_settings(..., json_settings=...) was never called with the taskfile dict. If you have divergent values between taskfile JSON and settings.ini, the taskfile value will now win — see CHANGELOG for the upgrade note.

Test plan

• pytest tests/unit/ — 732 passed, 5 skipped
• pytest tests/integration/ -m "not requires_tm1" — 36 passed
• ruff check src/ tests/ — clean
• Smoke: settings.ini with only default_tm1_instance set → DEPRECATION warning fires at load
• Smoke: settings.ini with both tm1_instance and default_tm1_instance → no warning
• Smoke: tasks push --target-tm1-instance X → still works, logs deprecation warning
• Smoke: tasks push --tm1-instance X → works, no warning
• TM1-backed integration tests (-m requires_tm1) — to be run by a maintainer with a live TM1

Notes

• Issue Detailed per-execution results for expandable tasks (--detailed-results) #146's literal-string regression assertion is rewritten as behavioural tests of resolve_tm1_instance; the # Issue #146 regression comments are preserved.
• Cube-sourced runs (--tm1-instance X --workflow Y) have no tier 2 because the cube schema doesn't store taskfile-level settings. This asymmetry is documented in docs/features/tm1-integration.md.

🤖 Generated with Claude Code