[RNE Rewrite] feat: add Kokoro TTS pipeline implementation

[IgorSwat]

Description

What this PR adds:

• Implementation of the Kokoro (text-to-speech) pipeline for one-way streaming (single input, streamed output)
• Implementation for boolean DType
• Demo app for Kokoro model

Introduces a breaking change?

• Yes
• No

Type of change

• Bug fix (change which fixes an issue)
• New feature (change which adds functionality)
• Documentation update (improves or adds clarity to existing documentation)
• Other (chores, tests, code style improvements etc.)

Tested on

• iOS
• Android

Testing instructions

Run the dedicated demo app.

Screenshots

Related issues

Checklist

• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• I have updated the documentation accordingly
• My changes generate no new warnings

Additional notes

The implementation of the Kokoro pipeline follows strictly the one already done for Supertonic. For this purpose, it got simplified a bit, with the following changes:

• Duration predictor no longer has 3 specialized input-size methods: instead of 'forward_32', 'forward_64' and 'forward_128' we use one 'forward' for the entire available token range. Costs a little bit worse performance, but gains reduced memory usage.
• Text partitioning and phonemization are applied in a simple, sequential order. In old Kokoro pipeline they would interleave themselves. In most practical cases it is a meaningless change, although if user passes a very long text in a language where Phonemis does not provide a lexicon (currently only English pipeline does), it can be significant for latency.

[msluszniak]

Example app looks solid, everything worked correctly. I will do a second pass on this one and I will be good with this PR :))

[barhanc]

Second pass. Tested example app and it works great. After the pending comments are resolved, I think we can merge this.