feat: resolve UUID page references in DB mode

[arndjan]

Summary

Fixes #21 — get_page_content in DB mode returns [[uuid]] instead of [[Page Name]] for page references, making output unreadable.

This PR adds automatic UUID-to-name resolution:

• Text format: [[69133208-...]] → [[Administratie]] (default behavior)
• JSON format: adds a resolved_refs mapping (uuid → name) alongside raw content, so agents can read names but use UUIDs for write operations
• resolve_refs parameter: opt-out with resolve_refs=false when raw UUIDs are needed

How it works

1. After fetching blocks, a regex (\[\[([0-9a-f-]{36})\]\]) extracts all UUID references
2. Unique UUIDs are batch-resolved via logseq.Editor.getPage, preferring originalName over name
3. A pre-built map is passed through _format_block_tree to replace refs during formatting

Design decisions

• Gated behind LOGSEQ_DB_MODE — markdown-mode users see zero change in behavior or performance
• Batch with dedup — duplicate UUIDs across blocks result in a single API call per unique UUID
• Graceful degradation — if a UUID can't be resolved (deleted page, API error), it's left as-is
• Resolution happens in run_tool, not _format_block_tree — collect all UUIDs first, resolve once, then pass the map down

Test plan

• 23 new tests covering:
  • UUID regex pattern matching (valid, invalid, bare UUIDs)
  • _collect_block_uuids (flat, nested, dedup)
  • _resolve_block_refs (replace, keep unresolved, mixed)
  • _format_block_tree with uuid_map (root, children, without map)
  • resolve_page_uuids API method (success, originalName preference, not found, dedup, API error, multiple)
  • Integration: resolve_refs enabled, disabled, markdown-mode skip, JSON format
• All 323 tests pass

🤖 Generated with Claude Code

[ergut]

Great work on this — the UUID resolution approach is clean and the batch dedup design is exactly right. We just merged PRs #31 and #33 which touched the same tools.py file, so this branch now has a conflict. Could you rebase on main? The changes don't overlap semantically so it should be straightforward. Thanks @arndjan!

[ergut]

Really well thought out, @arndjan. The batch dedup design is exactly right — collect all UUIDs first, resolve once, pass the map down. The resolve_refs opt-out and the resolved_refs mapping in JSON format are both thoughtful touches. 345 tests pass. Merging.