Generic markdown import (Obsidian vaults, knowledge bases)

## Problem

Users with existing knowledge in Obsidian vaults, Logseq graphs, or other markdown-based systems have no migration path into Palinode beyond manual copy-paste.

## Proposed solution

Generalize the existing OpenClaw migration parser into a generic markdown importer:

- Configurable heading-level splitting (`##`, `###`, etc.)
- LLM-powered extraction mode for unstructured brain dumps
- Frontmatter generation with category detection
- `--dry-run` and `--review` modes (already proven in OpenClaw importer)
- SHA-256 dedup to avoid re-importing existing content

**Target sources:** Obsidian, Logseq, Bear, plain markdown directories.

## Context

The OpenClaw importer (`palinode migrate openclaw`) already handles structured MEMORY.md files with scoring-based type detection. This extends that foundation to arbitrary markdown.