[PLT-0] Fix test expectations for updated test assets

[mrobers1982]

Description

Follow-up to #2055 (test-asset repath). The new hosted assets differ from the deleted originals, so several tests needed their fixtures/expected values realigned.

Changes

Bucket 1 — image_url byte-equality failures

• image_url pointed at https://picsum.photos/200/300.jpg, which returns a different random image per request, breaking every requests.get(image_url).content == requests.get(data_row.row_data).content assertion in test_data_rows.py. Repointed to a stable, byte-stable JPEG (potato.jpeg).
• This also fixes test_data_rows_upsert.py::test_multiple_chunks (assert 7 == 11): the shorter picsum URL under-filled the mocked 300-byte upsert chunks. Verified by reproducing the chunker locally — potato URL yields 11 upload calls, matching the assertion.
• Applied the same picsum → stable-JPEG swap in lbox-alignerr/tests/conftest.py (latent there — no byte-equality asserts — but removes the third-party dependency).

Bucket 2 — expected values for new asset content

• Text: TextEntity token updated to match sample-text.txt at offsets 112–128 (server extracts text[start:end+1]).
• Document: entity token IDs / groupId / text re-derived from the new sample-pdf-textlayer.json, and page width updated 612 → 594 (new test-pdf.pdf MediaBox).
• Video: frame_rate 10 → 25, frame_count 100 → 627 (intrinsic to test_video_500kb.mp4, verified via ffprobe).

Notes / not in scope

The following pre-date this change and reproduce on develop (backend/staging, not asset-related), so they are intentionally not addressed here:

• test_bulk_assign_data_rows* — project_based_user fixture uses the removed addMembersToOrganization mutation.
• test_get_slice — shared-org state-leak flakiness.

Verification

Test-only changes. Text/Document values are read directly from the new asset files (deterministic). The video frame_count = 627 and all server-rehost byte-equality checks can only be fully confirmed by the integration job running against staging.

🤖 Generated with Claude Code

---

Note

Low Risk
Test-only and dependency-pin changes; no production SDK logic beyond cosmetic formatting in project.py.

Overview
Follow-up to test-asset repathing: integration and export tests are updated so fixtures and assertions match the new hosted files, not the removed originals.

Stable image URL: IMG_URL in libs/labelbox and lbox-alignerr conftest switches from picsum.photos to a fixed GCS JPEG (potato.jpeg), with comments explaining byte-equality and upsert chunk-count tests that depended on a non-random source.

Annotation import expectations: Text entity token, document token IDs / groupId / selected text, and PDF page width (612 → 594) are refreshed for sample-text.txt, sample-pdf-textlayer.json, and test-pdf.pdf.

Video streamable export: Expected frame_rate (10 → 25) and frame_count (100 → 627) for test_video_500kb.mp4.

Dev deps: Root and lbox-alignerr pytest capped to <9 until pytest-cases supports pytest 9.

Minor formatting-only edits in project.get_overview, batch overview integration test, and unit tests—no API behavior change.

^{Reviewed by Cursor Bugbot for commit c778ba7. Bugbot is set up for automated code reviews on this repo. Configure here.}