Add PCM audio capture to browser audio input library

[mnemitz]

Problem

When recording audio in the browser, it's not very straightforward to capture PCM audio data, since the MediaRecorder API only provides encoded audio, and the scriptProcessor approach is both deprecated and flaky in terms of performance.

Solution

1. Update the existing @speechmatics/browser-audio-input package to expose
   • a PcmRecorder class
   • A minified PCM audio processor script, which can be run inside an AudioWorklet thread. See the MDN example here.
   • Updated README to explain how to consume the above script.
2. Update @speechmatics/browser-audio-input-react to include:
   • new hook usePcmAudioRecorder, to control a context-backed instance of the PcmRecorder class
   • new hook usePcmAudioListener, to subscribe to incoming PCM audio
   • Context provider PcmAudioRecorderProvider to provide the React context needed for the previous two
3. Update the NextJS example to use the above React bindings instead of the deprecated createScriptProcessor approach.

[anjz]

Looks great! Very nice explanation in the README's