[aisdk] add tokens count feature

[thiago-figueredo]

Summary

Token counting allows estimating the cost of a request and checking whether a prompt fits within the model's context window before making an API call.

• Add TokenCounter interface that LanguageModel implementations can optionally implement
• Add CountTokens and CountTokensStr convenience functions for counting tokens in messages
• Implement token counting for Anthropic provider using their Beta Messages CountTokens API
• Implement token counting for OpenAI provider using tiktoken (with proper encoding selection for different model families)
• Add mock support for token counting in tests

How was it tested?

• Added unit tests in count_tokens_test.go covering basic message counting, multiple messages, error handling, and unsupported models
• Added unit tests in tokenizer_test.go for OpenAI's tiktoken-based counting including encoding selection for different models
• Ran tests locally
• Ran aisdk/ai/examples/count_tokens/main.go

Community Contribution License

All community contributions in this pull request are licensed to the project maintainers under the terms of the Apache 2 License.

By creating this pull request I represent that I have the right to license the contributions to the project maintainers under the Apache 2 License as stated in the Community Contribution License.