feat: add MinerU document parsing SOP and helper

[GGzili]

What

Adds a MinerU document-parsing capability to memory/ as a SOP plus helper, matching the repo's tool-paired SOP pattern (e.g. procmem_scanner_sop.md + procmem_scanner.py):

• memory/mineru_sop.md: standard operating procedure (quick start, interface notes, limits)
• memory/mineru.py: zero-extra-dependency helper (uses requests). Submits a URL or local file to the MinerU API v4, polls, then downloads and extracts the resulting Markdown/JSON.
• .gitignore: whitelist entries for the two files.

Why

PDF / Office / image to Markdown+JSON parsing is a common, fundamental agent capability. This exposes MinerU (https://mineru.net) as a reusable core tool.

Testing

• python -m py_compile memory/mineru.py passes.
• Manual run: python memory/mineru.py <url|file> -o ./out (requires a MinerU token).