[WASM] Integration tests: Model providers

[awsarron]

Parent issue

strands-agents/sdk-typescript#1104

Scope

Write integration tests validating each model provider configuration:

• BedrockModel() - default provider, credential bridging from Python botocore
• AnthropicModel() - direct Anthropic API
• OpenAIModel() - OpenAI API (Chat Completions)
• GoogleModel() - Google GenAI (Gemini)
• Model parameters: max_tokens, temperature, top_p, stop_sequences
• System prompt configuration
• Model switching between invocations

Context

Model configs are Python factory functions that produce WIT model-config variant values. Bedrock is special: credentials are resolved on the Python side (botocore) because the WASM guest's browser-platform AWS SDK has no credential chain.

Key areas:

• Each provider's config serializes correctly through WIT
• Bedrock credential bridging works (access key, secret, session token, region)
• Model parameters affect generation behavior
• Provider-specific features (if any) work correctly

Done when

• BedrockModel works with default AWS credentials
• AnthropicModel works with API key
• OpenAIModel works with API key
• GoogleModel works with API key
• Model parameters (temperature, max_tokens) are respected
• System prompt is passed correctly to each provider
• Switching models between invocations works