refactor(core): honor canvas contexts throughout render passes

[ibgreen-openai]

Goal

Make deck.gl's existing render and picking passes consistently honor the active luma canvas context without introducing a multi-canvas API or changing the single-canvas default.

Changes

• Accept either a CanvasContext or PresentationContext in internal rendering and picking options.
• Propagate that context through synchronous/asynchronous point and rectangle picking, including the picking render pass.
• Derive framebuffer size, viewport coordinates, pixel ratio, and post-processing buffer sizes from the active context.
• Add focused coverage for context propagation and drawing-buffer/pixel-ratio behavior.

Review scope

Seven files; no new Deck props, canvas management, view-layout changes, widgets, or application examples. Omitting the context preserves the existing default behavior.

Validation

• yarn build
• Repository lint and formatting checks through the pre-commit hook
• yarn vitest run --project headless test/modules/core/lib/deck-picker.spec.ts test/modules/core/passes/layers-pass.spec.ts test/modules/core/passes/pick-layers-pass.spec.ts
• Node smoke tests through the pre-commit hook
• yarn test-headless --retry 1 --silent — 793 passed, 8 skipped; unrelated existing React-mount and LoadingWidget tests required their configured retry.

Stack

1. This PR: context-aware render/picking plumbing.
2. refactor(core): size and filter viewports by canvas #10475: a separate, two-file viewport-layout preparation PR.
3. Existing multi-canvas feature PR RFC/POC: Extend View system to support multi-canvas rendering #10229.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 99c2cbfdc2

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Clear views on the selected canvas framebuffer

When canvasContext is a non-default presentation context and a View has clear enabled, _render resolves that context's framebuffer here, but _drawLayers still forwards only the original target (typically null or undefined) and _drawLayersInViewport opens its clear pass with framebuffer: target. The clear consequently runs against the device's default canvas rather than the framebuffer used by the outer render pass, leaving stale color/depth in the selected canvas and potentially clearing another canvas; pass the resolved framebuffer into the per-view clear.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Resolve post-process output against the selected canvas

When a non-default canvasContext is rendered with a built-in PostProcessEffect and the canvas target is null, this only sizes the intermediate buffers from the selected context. _postRender still gives the final effect opts.target, so PostProcessEffect/ScreenPass starts its final pass with a null framebuffer, which selects the device's default canvas rather than canvasContext.getCurrentFramebuffer(); the selected canvas receives no composited output while the default canvas is overwritten.

Useful? React with 👍 / 👎.

[coveralls]

coverage: 83.374% (+0.002%) from 83.372% — codex/multi-canvas-context-plumbing into master