feat(pdf): merge paragraph fragments split across column/page breaks

docling joins a paragraph whose previous fragment ends mid-sentence (a letter,
not sentence punctuation) with a lowercase continuation (`…definition of` +
`lists in…` → `…definition of lists in…`); my pipeline emitted two paragraphs.
Add a post-assembly pass that merges consecutive top-level paragraph nodes on that
rule. A heading/table/figure between fragments ends the paragraph.

2203 277→255, 2305v1 68→56, with smaller gains on 2206/redp5110/normal_4pages;
the 4 exact PDFs are unchanged (their paragraphs end with sentence punctuation, so
nothing false-merges).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: artiz

crates/fleischwolf-pdf/src/assemble.rs

@@ -587,6 +587,39 @@ pub fn assemble_page(
     }
 }
 
+/// Merge paragraph fragments split across a column or page break. docling joins a
+/// paragraph whose previous fragment ends mid-sentence (a letter, not sentence
+/// punctuation) with a lowercase continuation: `…definition of` + `lists in…` →
+/// `…definition of lists in…`. Only consecutive top-level paragraphs merge — a
+/// heading/table/figure between them ends the paragraph.
+pub(crate) fn merge_continuations(nodes: &mut Vec<Node>) {
+    let mut i = 0;
+    while i + 1 < nodes.len() {
+        let merged = match (&nodes[i], &nodes[i + 1]) {
+            (Node::Paragraph { text: a }, Node::Paragraph { text: b }) => {
+                let a_open = a
+                    .trim_end()
+                    .chars()
+                    .next_back()
+                    .is_some_and(|c| c.is_alphabetic());
+                let b_cont = b
+                    .trim_start()
+                    .chars()
+                    .next()
+                    .is_some_and(|c| c.is_lowercase());
+                (a_open && b_cont).then(|| format!("{} {}", a.trim_end(), b.trim_start()))
+            }
+            _ => None,
+        };
+        if let Some(text) = merged {
+            nodes[i] = Node::Paragraph { text };
+            nodes.remove(i + 1);
+        } else {
+            i += 1;
+        }
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use super::clean_text;

crates/fleischwolf-pdf/src/lib.rs

@@ -108,6 +108,7 @@ impl Pipeline {
         pdfium_backend::for_each_page(bytes, password, |n, _total, mut page| {
             self.process_one_page(n, &mut page, &mut doc)
         })?;
+        assemble::merge_continuations(&mut doc.nodes);
         Ok(doc)
     }
 
@@ -187,6 +188,7 @@ impl Pipeline {
         for (n, page) in pages.iter_mut().enumerate() {
             self.process_one_page(n, page, &mut doc)?;
         }
+        assemble::merge_continuations(&mut doc.nodes);
         Ok(doc)
     }
 }