Feat/streaming config

[wenatuhs]

This PR adds support to configure streaming behavior on Osprey UIs through config.yml:

cli:
  # Streaming display configuration
  # Controls how LLM streaming output is displayed per node
  # Options: show | disabled  (CLI has no collapse, so "hide" maps to "disabled")
  streaming:
    respond: show                   # Final response (show | disabled)
    python_code_generator: disabled # Code generation (show | disabled)

tui:
  # Streaming display configuration
  # Controls how LLM streaming output is displayed per node
  # Options: show | hide | disabled
  #   show     - expanded during streaming, auto-collapses after completion
  #   hide     - collapsed from the start, user can expand to watch live
  #   disabled - not displayed at all
  streaming:
    respond: show                   # Final response (show | disabled; hide not allowed)
    python_code_generator: hide     # Code generation collapsed by default

openwebui:
  # Streaming display configuration
  # Controls how LLM streaming output is displayed per node
  # Options: show | hide | disabled
  #   show     - tokens stream live, code block visible
  #   hide     - buffered during streaming, collapsed <details> block after completion
  #   disabled - not displayed at all
  streaming:
    respond: show                   # Final response (show | disabled; hide not allowed)
    python_code_generator: hide     # Code generation collapsed by default

The above config shows the default behavior for CLI, TUI, and Open WebUI -- for respond the streaming is enabled, and for python code generator the streaming is hidden or disabled to not disturb users.

You don't have to change anything in the current config.yml of your agent to get the default behavior, but if you'd like to, say, completely disable the python code generator streaming (no code would show at all) in Open WebUI, you can put this snippet in your config.yml:

openwebui:
  streaming:
    python_code_generator: disabled