test(2meet-data-optimizer): add live test runner and failing shadow_read write suite

- Add tests/live: a `wp eval-file` runner (PASS/FAIL/XFAIL/XPASS/SKIP) with
  fixtures that override entity modes in process (pre_option) and read
  native / flat values with plain SQL, so WP core meta flows run for real
- shadow_write suite reproduces the dev36 finding: in shadow_read,
  update_*_meta returns false and native meta is never written
  (3 XFAIL marked for PR-056); the dual_write control passes
- Move the dev36 integration harness into tests/live/tools; outputs go to
  cwd / $TMDO_LIVE_OUT, site URL and path are env-configurable
- Verify: wp --path=<dev36> eval-file tests/live/run.php shadow_write
  → 1 passed, 3 xfail; unit 658 OK, integration 480 OK,
  phpcs (phpcs.xml) 0, stan OK
- PR: PR-052

tests/live/README.md

@@ -0,0 +1,55 @@
+# Live test suite
+
+Tests that must run inside a real WordPress site, because the stubbed
+`tests/integration` bootstrap cannot reproduce:
+
+- WP core meta flows (`update_metadata()` reads the old value through
+  `get_metadata_raw()` before writing — the shadow_read write-loss bug lives there)
+- the SQL that `WP_Query` actually builds, and hook priorities
+- HivePress / WooCommerce behaviour (`pre_get_posts` sort clauses, empty-key
+  meta reads, direct-SQL stock updates)
+
+**Never run against a production site.** Suites create and delete fixtures
+(`tmdo_live_*` users, `tmdo-live-*` terms). Entity modes are overridden in
+process through `pre_option_wpdo_bridge_modes`; nothing is written to
+`wpdo_bridge_modes` or `wpdo_hook_bus_enabled`.
+
+The reference site is dev36 (HivePress + Marketplace + WooCommerce), with this
+plugin symlinked from the `plugins-dev` worktree while developing.
+
+## Run
+
+```bash
+wp --path=/var/www/Studio/wp-local-dev36 eval-file tests/live/run.php <suite> [<suite> ...]
+```
+
+Exit code 0 when every test passed, 1 otherwise. Output per test:
+
+| Status | Meaning |
+|---|---|
+| PASS / FAIL | normal |
+| XFAIL | known bug, test marked `xfail` with the PR that fixes it; not counted as a failure |
+| XPASS | an `xfail` test passed — counted as a failure so the fixing PR removes the marker |
+| SKIP | the site does not register the key / taxonomy the test needs |
+
+## Suites
+
+| Suite | Covers |
+|---|---|
+| `shadow_write` | writes in shadow_read reach native meta (xfail until PR-056) |
+
+## Tools (`tools/`)
+
+Benchmark and data-flow helpers used for the 2026-09-26 dev36 integration test.
+Output files go to the current directory (`qbench`) or `$TMDO_LIVE_OUT`
+(default `/tmp`, `run_batch.sh`), never into the repo.
+
+| Tool | Use |
+|---|---|
+| `seed.php` | `wp eval-file seed.php <vendors\|listings\|requests\|reviews\|products\|orders\|terms> <n> [tag]` — seed through real HivePress / WC APIs |
+| `run_batch.sh` | `run_batch.sh <tag> [divisor]` — parallel seeding batch (`TMDO_LIVE_WP_PATH`, `TMDO_LIVE_OUT`) |
+| `clone.sql` | bench only: clone hp_listing rows to reach large scale (`sed "s/__K__/<n>/" clone.sql \| wp db query`) |
+| `checker.php` | `wp eval-file checker.php [entity]` — native vs flat consistency per registered key (SQL only) |
+| `qbench.php` / `qdiff.php` | run 16 WP_Query shapes, then diff two runs (native vs routed): result sets, order, latency, SQL errors |
+| `mut.php` | data-flow mutations through HivePress models, WC CRUD and core APIs; prints native / flat / API values |
+| `fe.sh` + `zz-s36-probe.php` | front-end search probe (`TMDO_LIVE_URL`). Copy the probe into `wp-content/mu-plugins/` only for the run and remove it afterwards; it only answers requests carrying `X-S36-Probe: 1` |