Un-xfail 9 im.tincan.Calls contract tests; make the pending-iface guard self-healing

[quad341]

What this changes

Nine D-Bus contract tests for the im.tincan.Calls interface (signals like
IncomingCall, CallConnected, CallEnded, CallActive, etc.) were
xfailing under a "not yet exported by the daemon" guard. The daemon has
actually exported all nine signals for a while now — the guard just never
got updated, so the tests were silently skipping real coverage instead of
running it.

This removes the stale entry so those 9 tests run for real (77 passed, 1
skipped, 0 xfailed in the contract file — was 66/2/10 before).

It also fixes why the guard went stale in the first place: the
pending-iface skip was an unconditional pytest.xfail(), which can never
self-heal or flag when an interface lands. It's now a strict xfail
(mirrors the existing _KNOWN_BROKEN_CONTRACT pattern in conftest.py), so
the next time an interface listed as "pending" actually ships, CI fails
loudly (XPASS) instead of the test suite quietly losing coverage again.

Review notes

• Test-only change — one file, tests/tincand/test_dbus_contract.py, no
  runtime/production code touched.
• The self-healing behavior was verified directly, not just read: a stale
  pending-iface entry was temporarily simulated and confirmed to produce a
  hard XPASS(strict) failure, then reverted before this commit.

Test plan

• pytest tests/tincand/test_dbus_contract.py — 77 passed, 1 skipped, 0 xfailed
• Full suite pytest tests/ — 2376 passed, 1 skipped, 0 failed, 0 xfailed
• ruff check . — 77 errors, identical count before/after (zero new lint issues)
• Release gate: release-gates/unxfail-calls-contract-sgeqs-gate.md

🤖 Deployed by actual-factory

[quad341]

Thanks @quad341 — this is exactly the kind of stale-guard cleanup that's easy to leave rotting for months. Nice catch that the im.tincan.Calls xfail was silently eating real coverage after the daemon started exporting all nine signals. What I especially like is that you didn't just delete the entry: converting the guard to a strict xfail (mirroring _KNOWN_BROKEN_CONTRACT in conftest.py) means the next "pending" interface that lands trips CI with a loud XPASS instead of quietly losing coverage again — that's the fix that keeps the fix from being needed twice.

I re-traced the core claim against tincand/dbus_service.py and confirmed all nine Calls signals are unconditionally exported, so those 9 §1 cases are genuinely running now. CI's green, it's test-only, and the change is self-contained — merging as-is, nothing needed on my end. Appreciate the thorough test plan and that you actually simulated the strict-xfail self-heal rather than just asserting it. 🚀

---

Reviewers: Qwen (local) — ok · Claude (claude-opus-4-8[1m]) — ok · Codex (gpt-5.5) — ok
Synthesis: claude-opus-4-8[1m]

Verified locally + tests pass. Ready to merge — clicking the button is the only step left.