Various updates to PsiStudio, Audio and Sigma

Fixed issues with moving to the previous step.
Added support for using specific synthesis voices via the Sigma client configuration.
Updated the Sigma app to use caching for speech synthesis.

Added SpeechSynthesisCache component to Microsoft.Psi.Speech to implement a cache for generated speech synthesis content.
The Azure-based SpeechSynthesizer component from Microsoft.Psi.CognitiveServices.Speech can now cache the generated utterances for future use (future identical requests do not need to go to the cloud, speeding up synthesis). The developer can also now select (via configuration) whether the audio buffers are streamed in real-time or as they arrive, and how close ahead of real time the generated audio buffers are streamed. These options allow for better controlling playback and avoiding drops.

Added a Streamline operator for audio streams that allows for normalizing gaps and overlaps in the audio buffer stream based on different methods (Concatenate, Pleat, Unpleat)
Added a batch processing task for exporting audio streams to wav files

Improved robustness of audio playback and addressed a number of issues that created drift between audio and visual playback.
Run batch task processing menu is now split into submenus by batch task processing namespace.
Added ability to export audio streams (or selections thereof) to a wav file.

Added a SessionImporter.OpenStream overload that allows for providing a [PartitionName]:StreamName stream specification. This simplifies the configuration of batch processing tasks where input streams need to be specified, eliminating the need for specifying the partition separately. The existing batch tasks were adjusted to leverage this feature.

Introduced Merge interval operator to compute a non-overlapping set of intervals that covers a given set of (potentially overlapping intervals)

microsoft / psi