Reject a ParallaxGroup depth of 0 when it's set

[pvcraven]

Summary

Fixes the first item of #1551: a ParallaxGroup layer with a depth of 0.

arcade.future.background.ParallaxGroup offsets each layer by the group's offset divided by the layer's depth. A depth of 0 was accepted when set, then raised ZeroDivisionError the first time the offset changed. That's usually in on_update, far from where the depth was set.

The fix

• Depth 0 is rejected when it's set: setting it now raises:

  A ParallaxGroup depth can't be 0. For a layer that doesn't scroll, use float('inf').

  The check covers all four places a depth can be set: the constructor, add(), change_depth() and group[i] = depth. A rejected depth isn't stored, so the group keeps working.

• Why not make depth 0 mean "doesn't scroll": layers scroll faster as their depth gets smaller, so 0 behaving like infinity would be surprising.

• Pinning a layer already works: infinite depth gives an offset of 0. The class docstring now says so. That covers the INFINITE_DEPTH item in the issue without a new constant.

Not in this PR

#1551 also asks for an initial offset per layer and a helper to sync the offset from a camera. Those are enhancements, so the issue stays open for them.

Tests

New tests/unit/test_background_parallax_group.py; ParallaxGroup had no tests before:

• The offset is divided by each layer's depth.
• Infinite depth doesn't scroll.
• Depth 0 is rejected in all four places, and nothing is stored when it is. This test fails on development.

The 5 background examples run. Full suite: 1759 passed. ruff, mypy and pyright all pass.

The changelog has an entry under Unreleased → Fixes.

🤖 Generated with Claude Code