[BUG] testinfra init fixture times out on uv sync of heavy extras (30s 'completed' wait)

[c-vigo]

Description

The session-scoped initialized_workspace fixture in tests/conftest.py times
out during the syncing_deps → completed transition, causing an ERROR at setup that cascades to every test depending on it (TestDevContainerStructure,
etc.).

The init pipeline reaches Syncing dependencies... (emitted by
assets/init-workspace.sh:398), which runs just sync →
uv sync --all-extras --all-groups (assets/workspace/justfile.project:59-60).
The test-project's optional dependencies are heavy scientific packages
(numpy==2.5.0, scipy==1.18.0, pandas==3.0.3, matplotlib==3.11.0,
jupyter==1.1.1/jupyterlab), totalling tens of MB. The image ships no warm
uv cache, so on a cold/slow network the download is still in flight when the
30s timeout for "Workspace initialized successfully" fires
(tests/conftest.py:350).

The captured before buffer shows the failure mid-download:

⠹ Preparing packages... (114/119)
pandas      ------ 10.22 MiB/10.41 MiB
matplotlib  ------ 10.08 MiB/10.61 MiB
numpy       ------ 10.32 MiB/15.88 MiB
jupyterlab  ------ 10.09 MiB/16.35 MiB
scipy       ------ 10.34 MiB/33.65 MiB

Current stage: syncing_deps
Time in stage: 30.0s (timeout: 30s)
✗ completed: not reached

This is a flaky-by-network failure: it passes when PyPI is fast or the uv cache
is warm, and fails on a cold/slow run. It is infrastructure flakiness, not a
product defect in the init script itself.

Steps to Reproduce

1. Cold uv cache (fresh image / no prior sync), unthrottled-but-realistic network.
2. Run the testinfra suite, e.g. just test / pytest tests/ against
   ghcr.io/vig-os/devcontainer:dev.
3. The initialized_workspace fixture setup runs the interactive init.
4. Observe ERROR at setup of ... with pexpect.exceptions.TIMEOUT at the
   "Workspace initialized successfully" wait.

Expected Behavior

The init fixture completes deterministically regardless of how long
uv sync spends downloading the test-project's heavy extras, and dependent
tests run.

Actual Behavior

The 30s completed timeout fires while uv sync is still downloading
numpy/scipy/pandas/matplotlib/jupyterlab, the session fixture errors, and all
dependent tests are reported as ERROR at setup.

Environment

• OS: Linux
• Container Runtime: Podman 5.8.2 (Nix store path)
• Image Version/Tag: ghcr.io/vig-os/devcontainer:dev
• Architecture: AMD64
• Python: CPython 3.14

Additional Context

• Relevant code:
  • tests/conftest.py:346-351 — stage_patterns_after_copy, completed timeout = 30s
  • tests/conftest.py:442-474 — the timeout/pytest.fail path
  • assets/init-workspace.sh:398-402 — Syncing dependencies... → just sync
  • assets/workspace/justfile.project:59-60 — uv sync --all-extras --all-groups
  • assets/workspace/pyproject.toml — heavy optional deps
• Because initialized_workspace is session-scoped, one slow sync errors the
  whole suite, matching the "many tests fail like this" symptom.
• Closely related to T2.2 — Make the testinfra suite portable (TEST ADAPTATION) #635 (make the testinfra suite portable / TEST ADAPTATION).

Possible Solution

One or more of:

1. Raise / make configurable the completed (and syncing_deps) timeout in
   tests/conftest.py to cover a cold-cache download (e.g. env-overridable, with
   a generous default like 180–300s). Lowest-effort fix.
2. Warm the uv cache in the image build (prefetch the test-project extras
   into UV_CACHE_DIR) so uv sync is offline-fast and deterministic — aligns
   with the Nix-built-image work.
3. Trim the test-project's --all-extras for the init smoke path, or sync
   --no-install-project/offline, so the structure tests don't pay for the full
   scientific stack.

Prefer (1) for an immediate unblock; (2) for the durable fix under the Nix image.

Changelog Category

No changelog needed (test infrastructure).