One Anthropic-style extended-thinking content block. Anthropic’s API
returns thinking output as a sequence of {type:"thinking", thinking, signature} blocks; the signature is a server-issued cryptographic
token that we MUST echo back unchanged on every subsequent assistant
turn or the API rejects the request with 400 The content[].thinking in the thinking mode must be passed back to the API. Per Anthropic
docs, thinking blocks must also appear before text/tool_use blocks
inside the assistant message — provider/claude.rs::format_messages
enforces that ordering.