feat(doc): add StructuredTable and PageInfo to ParsedDocument for stable machine-readable output

Phase 1: Add `tables[]` and `pages[]` output to ParsedDocument for stable
machine-readable consumption. StructuredTable includes row_count, column_count,
headers, and rows. PageInfo includes page number, block_count, and labels.

Phase 2: Use lopdf for position-aware PDF text extraction instead of
pdf-extract, enabling better text ordering and layout detection.

- Add StructuredTable and PageInfo types to doc/model.rs
- Add extract_tables_for_output() to consume_parse.rs
- Add parse_pdf_with_lopdf() for position-aware PDF extraction
- Add lopdf dependency for PDF parsing

Author: RoyLin

Cargo.lock

@@ -1,7 +1,7 @@
 [workspace]
 resolver = "2"
 members = [
-    "core"
+    "core",
 ]
 # Native SDK crates require special toolchains (Python/Node.js)
 # Build them individually: cargo build -p a3s-code-node / a3s-code-py