Older Safari: Handle speech sample rate errors, reuse audio context (#111)

- Workaround for #110 where we've observed that the sample rate of 19000 isn't supported. We'll aim to increase the sample rate in a separate PR but this will at least mean the simulator isn't put in a bad state.
- Reuse the audio context created after the user interaction, as there are limits to how many you can create on Safari 13 (4 in my BrowserStack testing). I think we always intended to do this, but careful review needed as this is a more significant change than I was anticipating. We already had clean up code for the nodes when the board stops. We never cleaned up the context. So I think we were assuming it lived forever. If this doesn't work out then an alternative is to try closing the old one before creating the new one.

Author: microbit-matt-hillsdon

src/board/audio/index.ts

@@ -67,10 +67,12 @@ export class Audio {
   }
 
   async createAudioContextFromUserInteraction(): Promise<void> {
-    this.context = new (window.AudioContext || window.webkitAudioContext)({
-      // The highest rate is the sound expression synth.
-      sampleRate: 44100,
-    });
+    this.context =
+      this.context ??
+      new (window.AudioContext || window.webkitAudioContext)({
+        // The highest rate is the sound expression synth.
+        sampleRate: 44100,
+      });
     if (this.context.state === "suspended") {
       return this.context.resume();
     }

src/jshal.js

@@ -240,14 +240,22 @@ mergeInto(LibraryManager.library, {
     /** @type {number} */ buf,
     /** @type {number} */ num_samples
   ) {
+    /** @type {AudioBuffer | undefined} */ let webAudioBuffer;
+    try {
+      // @ts-expect-error
+      webAudioBuffer = Module.board.audio.speech.createBuffer(num_samples);
+    } catch (e) {
+      // Swallow error on older Safari to keep the sim in a good state.
+      // @ts-expect-error
+      if (e.name === "NotSupportedError") {
+        return;
+      } else {
+        throw e;
+      }
+    }
     // @ts-expect-error
     Module.board.audio.speech.writeData(
-      Module.conversions.convertAudioBuffer(
-        Module.HEAPU8,
-        buf,
-        // @ts-expect-error
-        Module.board.audio.speech.createBuffer(num_samples)
-      )
+      Module.conversions.convertAudioBuffer(Module.HEAPU8, buf, webAudioBuffer)
     );
   },