chore: Upgrade to luma.gl 9.3.0-alpha.10

[ibgreen]

Summary

This PR upgrades deck.gl to @luma.gl/* 9.3.0-alpha.10 and updates the repo for the latest luma shader/block layout behavior, Node 22/Vite behavior, and website bundling changes.

What Changed

luma.gl alpha.10 upgrade

• Upgraded workspace @luma.gl/* dependencies to 9.3.0-alpha.10
• Regenerated yarn.lock
• Updated the luma core API usage in modules/core/src/lib/attribute/gl-utils.ts to use dataTypeDecoder.getDataType(...)

Shader / uniform block compatibility

• Added layout(std140) to deck GLSL uniform blocks across modules
• Reverted the earlier temporary vec4/.xyz project-uniform workaround after the luma fixes landed
• Updated WGSL uniform declarations where needed for current luma expectations
• Kept the data filter module on a stable, explicit block layout to avoid offset drift between JS-packed uniforms and shader reads

DataFilter fixes

• Reworked modules/extensions/src/data-filter/shader-module.ts so the dataFilterUniforms block has a fixed layout with padded vec4 numeric ranges and an explicit uvec4 categoryBitMask
• Added shader-side accessors so scalar/vec2/vec3 filter paths still read the correct subset of those packed values
• Updated the JS-side uniform payload generation to pad numeric filter ranges consistently
• Updated unit tests to assert the packed UBO representation now stored in luma’s uniform store

Resource cleanup fixes

• Fixed ScenegraphLayer teardown to destroy tracked GLTF materials as well as scenegraph nodes
• Added explicit Tile3DLayer tileset destruction during finalizeState

Node 22 / Vite compatibility

• Added vite.config.js and vite.config.mjs to strip with {type: 'json'} import attributes during Vite/esbuild processing
• This keeps the repo test/build flow working under Node 22

Website build fixes

• Moved the deprecated Docusaurus onBrokenMarkdownLinks setting to markdown.hooks.onBrokenMarkdownLinks
• Added a website webpack alias for @luma.gl/webgl/constants to work around the alpha.10 package resolution issue during docs bundling

Validation

• yarn build passes
• yarn test no longer has the DataFilterExtension / GeoJsonLayer#DataFilterExtensionWithBinaryAttributes failures
• Current remaining yarn test failures are render snapshot diffs only:
  • arc-lnglat
  • arc-lnglat-wrap-longitude
  • path-globe
  • icon-lnglat-facing-up

Notes

• The website/Docusaurus build no longer hits the previous @luma.gl/webgl/constants resolution error locally
• I was not able to complete website/yarn test-build end-to-end on this machine because the local Docusaurus build was later killed with exit 137

[coveralls]

coverage: 79.906% (+0.03%) from 79.878%
when pulling b4fdf87 on ib/luma-930-alpha.9
into e87b04d on master.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 11f29b4278

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]

Pad ProjectUniforms before opting it into std140

Making projectUniforms a layout(std140) block changes every vec3 in this UBO to a 16-byte slot, but the matching WGSL struct in modules/core/src/shaderlib/project/project.wgsl.ts still packs projectionMode, devicePixelRatio, commonOrigin, etc. immediately after the preceding vec3<f32> fields. On WebGPU, any layer that uses the shared project module will therefore read those uniforms from the wrong offsets and produce corrupted transforms. This needs explicit padding on the WGSL side (or the GLSL block must stay out of std140).

Useful? React with 👍 / 👎.

[ibgreen]

Why? Doesn't luma.gl calculate offsets using the appropriate layout for each device?

[Pessimistress]

What is the purpose of these root level configs?