Adding tool streaming output via Flux to support LLM invoking or Agent invoking #57

I have a scene in which my MCP client calls my MCP server that invoking LLM SSE API,
I assume that my client could consume stream chunks instead of waiting until all chunks produced.

But I found interfaces in this SDK just support normal sync request or async request.

I wish it have a Flux response tool calling style, and I'm willing to contribute.

Mono<McpSchema.CallToolResult> callTool(McpSchema.CallToolRequest callToolRequest);

Flux<McpSchema.CallToolResult> callStreamTool(McpSchema.CallToolRequest callToolRequest);