Refactor PCM audio utilities and add support for visualizers

[mnemitz]

Package changes

⚠️ This is a breaking change to the following packages:

• @speechmatics/browser-audio-input
  • Update PCMRecorder class to clean up the AudioWorkletNode, rather than assuming the AudioContext will be closed.
  • Automatically resume audio context when startRecording is called, throwing a specific error of this fails.
• @speechmatics/browser-audio-input-react
  • Split into two hooks: usePCMAudioRecorder and usePCMAudioRecorderContext (first renders inline, second uses the first as a value to a context provider)
  • AudioContext must be supplied to usePCMAudioRecorder or PCMAudioRecorderProvider, rather than passed to startRecording
• @speechmatics/web-pcm-player-react
  • Supply extra hook usePCMAudioPlayerContext, plus context and provider for audio player (useful for sharing player across the app)
  • Audio context must be supplied to either usePCMAudioPlayer() or PCMAudioPlayerProvider

Examples changes

NextJS Flow:

• Add audio visualizer component, use for both agent and speaker transcript blocks
• Update dependencies, including React types (which were mistakenly at 18 instead of 19)

NextJS real-time

• Update dependencies, including React types (which were mistakenly at 18 instead of 19)

[TudorCRL]

SDK code lgtm! I think there might be an issue with the useHydrated hook in the examples which means that a valid audioContext is never returned. Maybe we can have a chat about this offline?