Improve Microsoft.ML.Tokenizers and drive complete 1.0 API

[ericstj]

Goal: Enable .NET developers to use tokenizers in their data pre-processing pipelines as part of their embedding and token generation tasks using language models.

Committed:

• [ ] Add support for more commonly used Tokenizers
  • [x] TikToken https://github.com/dotnet/machinelearning/pull/6981
  • [x] LlamaTokenizer & SentencePiece algorithm https://github.com/dotnet/machinelearning/issues/6987
  • [x] CodeGenTokenizer & Byte-level BPE https://github.com/dotnet/machinelearning/issues/6992
  • [x] WordPiece algorithm https://github.com/dotnet/machinelearning/issues/6988
  • [x] BERTTokenizer https://github.com/dotnet/machinelearning/issues/6991
• [x] Measure and improve performance of Tokenizers API - making breaking changes where necessary. (https://github.com/dotnet/machinelearning/issues/6982)
• [ ] Explore existing construction patterns to improve usability - both in factory API and load from configuration.
• [x] Drive adoption of Microsoft.ML.Tokenizers in other libraries
• [ ] Docs and samples

Backlog:

• [ ] Investigate using Microsoft.ML.Tokenizers in Azure OpenAI SDK
• [ ] Sentencepiece Unigram https://github.com/dotnet/machinelearning/issues/7186
• [ ] CLIP Tokenizer https://github.com/dotnet/machinelearning/issues/6993