Fix macOS stock-level audio exhaustion

[themuffinator]

Summary

• build and bundle checksum-pinned OpenAL Soft 1.25.1 for macOS release, validation, sanitizer, and universal2 paths
• keep the runtime package-relative at Contents/Frameworks/libopenal.1.dylib, with architecture, deployment-floor, dependency, signing, archive, and source/licence validation
• make OpenAL sample-buffer allocation/upload failures non-fatal so a successfully loaded map is not torn down
• retain Apple's OpenAL framework as a compile-only compatibility corridor
• update player, build, support-policy, planning, attribution, and candidate release documentation

Why

Issue #122 provided strong evidence that the published 0.12.0 package was still using Apple's deprecated OpenAL framework and exhausting its buffer pool on stock single-player maps. Lowering s_maxSoundsPerShader only moved the threshold, while s_noSound 1 allowed the same maps to stay loaded.

Bundled OpenAL Soft is the primary fix. The non-fatal allocation path is defense in depth for resource exhaustion or unusual third-party providers.

Dependency and licence review

• upstream: OpenAL Soft 1.25.1
• source: official GitHub tag archive
• verified SHA-256: 5f8efe8dfba5e9307a50251ba615ace857c7fa9dddfe34130b83e213d7f7cf24
• linkage: dynamic
• licence: LGPL v2 or later; compatible with this GPLv3 repository
• package payload: LGPL text, source/build notice, PFFFT/fmt/Microsoft GSL notices, and exact corresponding source archive

Validation

• full Windows x64 engine, dedicated server, SP/MP GameLibs, OpenGL renderer, and Vulkan renderer build
• macos_openal_provider_policy.py
• macos_metal_bridge.py
• macos_universal2_assembly.py
• macos_sanitizer_ci.py
• release_tooling_safety.py
• documentation-link and source-charset checks
• shell and Python syntax checks
• actionlint on all changed workflows
• broad macos-static run: every Single-player levels abort back to the main menu: CreateOpenALBuffer fails on Apple's OpenAL (macOS arm64, 0.12.0) #122 provider/package test passed; it also exposed an unrelated texture-filter push-smoke wiring regression on current main, which will be corrected separately

Physical Apple-hardware retesting is still needed before strengthening the project's experimental macOS support claim.

Fixes #122

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f9ecccc85e

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".

[themuffinator]

CI triage update: I compared the four red macOS Push Verification jobs with the current main-branch run at 0751161. Main fails the same renderer matrix cases with the same outcomes (foundation/PBR exit -11; cluster/default renderer expectations), while this branch completes the OpenAL Soft preparation, build, tests, install, and payload upload. The matching main run is https://github.com/themuffinator/openQ4/actions/runs/32992669468 and the branch run is https://github.com/themuffinator/openQ4/actions/runs/33002292512. I am therefore treating those four renderer-smoke failures as inherited baseline failures, not regressions from this audio change; the platform commit-validation jobs remain the merge gate.