chore(deps): scitex-clew ==0.17.0 -> >=0.20.0 — let the PostgreSQL migration arrive

[ywatanabe1989]

Moves scitex-clew from an exact pin to a floor, so the PostgreSQL store migration actually reaches this package.

Why this was the last SQLite item. scitex-python pinned scitex-clew==0.17.0 in three places. That tag is dated 2026-07-06, nearly two months before clew migrated its stores off the embedded engine (4f20c97 on 08-28, 416f42f/#151 on 08-30). So running this package's suite re-created .scitex/clew/runtime/clew.db — 84 KB, first sixteen bytes SQLite format 3 — even though clew's own develop had been clean for days. The exact pin froze a July build for two months and silently kept the retired engine alive.

The floor, not a bump. An exact pin is what caused this: it cannot pick up a fix without a human editing three lines. >=0.20.0 lets the migration arrive on its own, and is the shape the operator asked for when he raised it (「フロアをピンするだけのほうがいいのかなぁ」).

0.20.0 is verified at the ARTIFACT, not the version string — the trap that started this whole investigation was 0.19.1 looking newer than the work it lacked. Downloaded scitex_clew-0.20.0-py3-none-any.whl from PyPI and inspected it:

• _db/_core.py imports scitex_dev.store — yes
• _db/_core.py imports sqlite3 — no
• .py files importing the engine anywhere in the wheel — 0

Three call sites updated: lines 82, 796, 951.

[ywatanabe1989]

Superseded by #375 — please close this one; I am not touching this branch.

#375 does the same three-pin ==0.17.0 -> floor change plus the part that actually makes it work, which this PR is missing: scitex-dev==0.28.0 -> 0.57.0.

Raising clew's floor alone does not fix the red here. scitex-python pins scitex-dev EXACTLY, so the resolver ignores clew's floor and installs 0.28.0 regardless — which has no scitex_dev.store at all, so the install succeeds and the break lands at import (ModuleNotFoundError: No module named 'scitex_dev.store', three clew integration tests, all three Python legs).

#375 also raises the floor to >=0.20.1, because clew 0.20.0's own metadata was wrong (scitex-dev>=0.17.8, satisfied by 0.28.0 — that is why the bad pair resolved at all). scitex-ai/scitex-clew#158 fixes it to >=0.49.2, a floor measured by a real write-then-read rather than by importing: 0.43.1 is the first release carrying scitex_dev.store and host_store, imports every symbol clew imports, and still raises TypeError: tuple indices must be integers or slices, not str inside Store.rows(). Measured with the published clew wheel — 0.28.0 FAIL, 0.43.1 FAIL, 0.46.0 FAIL, 0.48.0 FAIL, 0.49.0 FAIL, 0.49.2 OK, 0.49.3 OK, 0.50.0 OK, 0.57.0 OK.

Sequencing: #375 stays unpushed until clew 0.20.1 is on PyPI, otherwise its own floor cannot resolve.