Pull test_prv_006/007 out of the default smoke gate (unresolved concurrent-load failures)

[saqibmanan]

Why

Both tests fail reproducibly under concurrent (`-n 3`) provider-smoke CI runs — confirmed on the post-#99-merge run on `CI`, 3/3 attempts including auto-reruns:

• `test_prv_006_org_create_dataset`: `TimeoutException: Timed out waiting for Metadata tab to appear after creating dataset` — a wait already bumped 30s→60s for this exact reason (2026-09-15), still failing.
• `test_prv_007_org_create_usecase`: `AssertionError: Summary mismatch. Expected: '...', Found: ''`

The org-collision theory that correctly explained (and fixed) `test_prv_011`'s flakiness in #99 does not hold here — `TEST_EMAIL_3` was moved off the org it shared with `TEST_EMAIL_2` specifically to test this, and both tests still failed identically on the next real run. Root cause is still open.

Given they're not safe to trust as a merge gate right now, pulling them out of the default `-m smoke` selection rather than leaving PR/push CI red (or silently green via more reruns masking a real, unexplained, consistent failure).

What changed

Removed `@pytest.mark.smoke` from both, kept `@pytest.mark.functional`. They still run on `workflow_dispatch` (`-m "smoke or functional"`) — not lost from coverage, just out of the default gate until root-caused.

Verification

Collection diff, confirmed directly:

• `-m smoke`: 9/18 provider/smoke tests collected (was 11/18) — 006 and 007 gone, 006b (unaffected, different test) still present.
• `-m "smoke or functional"`: 18/18 — both still collected and runnable on dispatch.