Important bugfixes to OpenAL implementation.

[geneotech]

• The audiobuf in offset -= src.bufQueue[src.bufsProcessed].audiobuf.duration; is misspelled - audioBuf was likely intended. I found this because I was getting a TypeError: src.bufQueue[src.bufsProcessed] is undefined, although I'm not 100% sure it was caused by the typo.
• switch cases for AL_SOURCE_DISTANCE_MODEL were checking against the string 'AL_SOURCE_DISTANCE_MODEL' instead of the integer value 0x200, causing AL_INVALID_ENUM on alIsEnabled, alEnable and alDisable with this enum.
• updateSourceTime sometimes leaves the src.bufOffset with a NaN value, although sadly I'm not sophisticated enough to understand why. I have added a harmless check that just sets it to 0 in case it turns NaN, which arguably shouldn't break anything, and it completely fixes the following crash:

  TypeError: AudioBufferSourceNode.start: Argument 2 is not a finite floating-point value.
      scheduleSourceAudio http://localhost:6931/Hypersomnia.js:8343
      setSourceState http://localhost:6931/Hypersomnia.js:8498
      sourceSeek http://localhost:6931/Hypersomnia.js:8853
      setSourceParam http://localhost:6931/Hypersomnia.js:9397
      _alSourcef http://localhost:6931/Hypersomnia.js:10089
      __emscripten_receive_on_main_thread_js http://localhost:6931/Hypersomnia.js:7102

  I have already battle-tested this one if clause in my browser game and it seems to work.

[geneotech]

Hello! Do you mean for AL_SOURCE_DISTANCE_MODEL? I'm afraid the other issues are too complex to be easily reproduced in a simple test.

[sbc100]

Hello! Do you mean for AL_SOURCE_DISTANCE_MODEL? I'm afraid the other issues are too complex to be easily reproduced in a simple test.

Whatever can be reasonably tested should be yes. Perhaps the first 2 parts of this change at least should be fairly straight forward to test?

[sbc100]

In fact, I wonder if we can split out the updateSourceTime into its own PR, since it seems a little more complex, and worthy of some deeper analysis, whereas the other two fixes look trivial.

For the updateSourceTime issue in particalar I'd be interested to know which of the assignments to src.bufOffset causes is used when its is NaN. Are you able to catch it in devtools? Can you see which of the three branches in updateSourceTime that assigned to bufOutput were taken?

[geneotech]

Thank you! This test to alDisable/alEnable/alIsEnabled should cover it, the panning still seems to work. As for the audiobuf typo, I'm really not sure how to cover it because the implications of the typo being there are really somewhat complex.

I'll make the separate PR with updateSourceTime only once I find the real cause.