Add streaming support for real-time AI responses

[jhs]

Overview

Enable streaming support for real-time token-by-token display of AI responses in both Python and TypeScript implementations.

Benefits

• Improved user experience with progressive response rendering
• Reduced perceived latency
• Better for long responses

Implementation Notes

• Must maintain isomorphic implementation across Python and TypeScript
• Should integrate with existing Interviewer.go() flow
• Consider how streaming interacts with field validation and tool calls

Scope

• Python implementation
• TypeScript implementation
• React hook integration (via onToken callback)
• Documentation and examples

Related

• Mentioned during React integration design discussion
• Would enhance useChatfield hook with streaming: true option