docs: Update streaming instructions for generative AI models

[Maki-Grz]

Documentation improvements for streaming:

• Clarified that streaming can be enabled for generative AI models to return responses incrementally, not just for Amazon Bedrock models.
• Specified that for Amazon Bedrock models, streaming is invoked by using the $DEPLOYMENT_URL/invoke-with-response-stream endpoint.
• Added instructions for enabling streaming with the Converse API by calling the $DEPLOYMENT_URL/converse-stream endpoint instead of /converse.

Ref: #79

[cla-assistant]

All committers have signed the CLA.