BATCH and MUTLI-SEQUENCE generation

[IgorAherne]

This PR adds efficient batch processing and multi-sequence generation.

Changes:

• Batch Generation: Process a list of different texts at once for a significant speed-up.

  model.generate(["Hello world.", "This is a test."])

• Multiple Variations: Use num_return_sequences to efficiently create diverse samples from the same text.

  model.generate("Hello world.", num_return_sequences=3)

The core inference pipeline has been updated to support these features in parallel, and example_tts.py is updated to demonstrate usage.