feat: add extensive tests and new tokenizers, pretokenizers, and decoders

[CodeWithKyrian]

What:

• Bug Fix
• New Feature

Description:

This PR introduces a comprehensive suite of tests for various tokenizers, alongside fixes for issues identified during the test development process. Additionally, several new pretokenizers, decoders, and pretrained tokenizers have been added, as detailed below:

New Additions

1. Pretokenizers:
   • WhitespacePretokenizer: Added support for tokenizing text by whitespace.
2. Decoders:
   • VitsDecoder: Introduced a decoder for improved text-to-speech decoding performance.
3. Pretrained Tokenizers:
   • BlenderBotTokenizer
   • CohereTokenizer
   • NougatTokenizer
   • GemmaTokenizer
   • Grok1Tokenizer
   • SiglipTokenizer
   • SpeechT5Tokenizer
   • VitsTokenizer

Modifications

• SplitPretokenizer: Corrected functionality to ensure proper behavior, addressing issues with its previous implementation.

Bug Fixes

During the development of the test suite, several bugs in the tokenizers were identified and fixed to enhance reliability and functionality.