Emscripten Updates from Jonathan Frank

[danoli3]

Combined Emscripten PRs, cherry-picked with original author (Jonathhhan) and adapted for 6.0.6:

• Emscripten glInfoExample fix #7211 glInfoExample fix (a72b1a1) — wrap GL limits in #ifndef TARGET_EMSCRIPTEN
• Emscripten audioDeviceSelect draft #7261 audioDeviceSelect (2501f2b) — add library_audioDeviceSelect.js, update library_html5audio.js, add --post-js
• Emscripten record video #7553 record video (dba81b4) — add recordVideo.js, fix-webm-duration.js, update html5audio (MediaStreamAudioDestinationNode), add --pre-js
• New Emscripten audioWorklet draft #7727 audioWorklet (0b26f61) — rewrite ofxEmscriptenSoundStream to audioWorklet, update html5audio.h, sound examples to 128 buffer, add USE_PTHREADS/AUDIO_WORKLET/WASM_WORKERS flags
• 6.0.6 bump (00f5336) — emsdk 5.0.7->6.0.6, template EMSCRIPTEN_MAJOR ifdef, install script fixes, FULL_ES2->FULL_ES3, conditional preload

[danoli3]

Dropping WASM_WORKERS in light of -s AUDIO_WORKLET=1 - fixes MAIN_MODULE issues

[danoli3]

Rebased on master and fixed: Emscripten audio now builds and plays

Rebase: the template typo fix is dropped, because it's already on master (#8559). The glInfoExample conflict with #8579 is resolved by keeping both the TARGET_EMSCRIPTEN guard and master's MSVC freopen_s path. Jonathan's commits keep their author.

New commit afa5f09, fixing why this didn't build (CI build-emscripten was failing) and why it made no sound once built:

1. MAIN_MODULE: emscripten rejects WASM_WORKERS with any MAIN_MODULE, so =2 didn't help. Removed it; EXPORT_ALL still exports every symbol for the JS libraries. MAIN_MODULE had been hiding the next three problems.
2. library_html5audio.js wasn't linked, so the html5audio_* functions were undefined. It's now passed as --js-library. It also now depends on $emAudio: emscripten 6.0.11's $emscriptenGetAudioObject reads emAudio without declaring that dependency.
3. EMBIND_AOT now defaults to 0, as the config comment already says. With no EMSCRIPTEN_BINDINGS (OF core has none) the link fails with expected to find pattern <<< EMBIND_AOT_INVOKERS >>>. Projects with their own bindings can still set it to 1.
4. WEBAUDIO_DEBUG=1 removed: its logging calls dbg(), which doesn't exist in Release builds (ReferenceError: dbg is not defined).
5. The audio context is back: ofxEmscriptenSoundPlayer's html5audio calls were disabled in cd463eeb6, so ofxEmscriptenAudioContext() compiled to context = initialized = true. The stream got audio context 1, which doesn't exist.
6. The --pre-js files now work in the audio worklet. recordVideo.js (drawTexture__deps = …, document.createElement) and fix-webm-duration.js (window.…) threw there, so the worklet processor never started. Errors in the worklet don't reach the page console; this one only showed up by attaching to the worklet target.

Tested: emsdk 6.0.11, headless Chrome on macOS, served with COOP/COEP headers.

Example	Build	Runtime
audioOutputExample	✅	context running, worklet processor running, output peak 0.05 (volume 0.1), no errors
soundBufferExample	✅	output peak 0.40, no errors
audioInputExample	✅	input callback running (3512 buffers in 10 s, fake mic)
soundPlayerExample	✅	context running, no errors (playback not exercised headless)
glInfoExample	✅	no errors

Still to look at:

• audioOutputExample plays its tone, but its waveform panes stay empty.
• Video recording wasn't exercised.
• Nothing was tested on a real browser with a microphone or speakers.

[danoli3]

The empty waveform panes in audioOutputExample aren't from this PR: open ofPolyline / ofBeginShape outlines don't draw on WebGL on master either. That's fixed separately in #8585, with before/after on WebGL, iOS and Android.

[danoli3]

Record video tested, with fixes in f4e02d8

Recording is started and stopped from the app the same way as in #7553: EM_ASM(recordVideo($0), isRecording);. For this test, a temporary r key toggle in graphicsExample, driven from headless Chrome with emsdk 6.0.11.

graphicsExample, 5 s recorded in the browser (VP9 WebM, 1024×768, 30 fps; shown here as a GIF):

Original recordings: graphicsExample-recordVideo.webm (video only) and audioOutputExample-recordVideo-with-audio.webm (VP9 + 44.1 kHz stereo PCM, the example's tone at about −23 dB peak).

Fixed (f4e02d8):

• Apps with no audio couldn't record. recordVideo always added AUDIO.contextStream's audio track, so graphicsExample failed with Cannot read properties of undefined (reading 'stream'), and the stop call then failed on the undefined MediaRecorder. The audio track is now added only when an audio context exists, and stopping when not recording does nothing.
• Recordings had no audio. The audioWorklet rewrite of library_html5audio.js dropped AUDIO.contextStream from the record video change. It's restored: a MediaStreamAudioDestinationNode fed by the analyser.
• Wrong file name. The download was always called Video.mp4, but Chrome records WebM. It's now named after the recorded type (Video.webm).

Notes:

• recordVideo.js is a --pre-js, so at -O3 the function is only kept in the build when the app references it from EM_ASM, as in the snippet above. Calling it from the browser console won't work.
• The ofTexture overload, recordVideo(1, textureID), wasn't tested.

[danoli3]

The 4 examples running (WebGL2, headless Chrome, emsdk 6.0.11)

Built with make from this PR merged with current master (f9e8d95, which includes #8585, so the audio waveforms draw). The mouse is moved in a circle over the canvas while recording.

Example	Audio	Errors
glInfoExample	n/a	0
audioOutputExample	context running, audioWorklet running, output peak 0.05 (volume 0.1)	0
audioInputExample	audioWorklet input, 2354 buffers received	0
soundBufferExample	output peak 0.47	0

The input example's microphone is a 330 Hz tone injected through getUserMedia, because headless Chrome's fake capture device is silent.

glInfoExample

audioOutputExample (mouse x pans, y sets the frequency)

audioInputExample

soundBufferExample