Diagnostics read the factory-owned engine-wide state and the renamed
platform (formerly built-in) device state: requested shows as one
optional request, the resolved/active split replaces tri-state
booleans, and the redundant has* flags are gone.

Per-effect mode types drive the pickers generically, so the high-pass
filter row only offers its legal modes. setAudioProcessingOptions now
returns a success-only result and throws on failure. Diagnostics use the
path-grouped component state and the renamed platformVoiceProcessingState,
and the voice processing toggle follows the platform-allowed rename.

The SDK dropped the flag from AudioProcessingState.

Pins the release branch revision ae672652 so LiveKitSDK.version reports
2.15.2. TestFlight build 2.15.2b20260717.1.

Demonstrates RoomOptions(defaultAudioCaptureOptions:), the way most apps
configure processing modes. The mic toggle still passes options
explicitly, which takes precedence and picks up panel changes made
after connecting.

The mode is irrelevant while the effect is disabled; graying the picker
makes that visible.

Default restores the communication profile via the AudioProcessingOptions
initializer; No processing uses the .noProcessing preset.