Add adaptive Neuroglancer pyramid policies

[akgohain]

What changed

• select Neuroglancer 2d/flat on-the-fly pyramids for materially anisotropic 3D data and 3d/isotropic pyramids otherwise
• apply the same deterministic policy to image and segmentation layers
• bound generated chunk size, downsampling amount, terminal size, and scale count
• preserve older Neuroglancer constructor compatibility through signature-aware keyword filtering
• record the selected policy in viewer responses, workflow metadata, and runtime events
• enable CI and Super-Linter for PRs whose base is an agent/** stack branch

Why

The volume pyramid contract in #176 supplies coordinate truth. This layer makes Neuroglancer use a bounded, anisotropy-aware rendering strategy today while later proofreading work adopts native level-aware reads and correction overlays.

Stack

Depends on #176 and is part of native stack #174.

Validation

• uv run pytest -q — 356 passed, 9 subtests passed
• focused volume/Neuroglancer suites — 43 passed
• direct construction against installed Neuroglancer 2.38
• Black and git diff --check pass