feat: add multimodal rawdata support

[starboyate]

Summary

• add pluggable multimodal rawdata ingestion support across the memind runtime
• introduce dedicated rawdata plugins and starters for document, image, audio, and tool-call inputs
• wire parser/resource handling into extraction flow so file, url, and direct-content inputs can be resolved consistently
• add vision-based image caption generation and multimodal processing support in server runtime and examples
• migrate related integration paths to Jackson 3 and clean up outdated implementation/spec docs

Test Plan

• mvn test