test: Add snapshot regression tests with PDF fixtures

Add 5 stripped public-domain PDF fixtures and golden markdown snapshots
for CI regression testing. Fix non-deterministic output caused by
HashMap iteration order in font stats, table heuristics, and rect
clustering by adding deterministic tie-breaking.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

src/markdown/analysis.rs

@@ -21,9 +21,12 @@ pub(crate) fn calculate_font_stats_from_items(items: &[TextItem]) -> FontStats {
         }
     }
 
+    // Break ties by preferring the smaller font size for deterministic output
     let most_common_size = size_counts
         .iter()
-        .max_by_key(|(_, count)| *count)
+        .max_by(|(size_a, count_a), (size_b, count_b)| {
+            count_a.cmp(count_b).then_with(|| size_b.cmp(size_a))
+        })
         .map(|(size, _)| *size as f32 / 10.0)
         .unwrap_or(12.0);
 
@@ -45,9 +48,12 @@ pub(crate) fn calculate_font_stats(lines: &[TextLine]) -> FontStats {
         }
     }
 
+    // Break ties by preferring the smaller font size for deterministic output
     let most_common_size = size_counts
         .iter()
-        .max_by_key(|(_, count)| *count)
+        .max_by(|(size_a, count_a), (size_b, count_b)| {
+            count_a.cmp(count_b).then_with(|| size_b.cmp(size_a))
+        })
         .map(|(size, _)| *size as f32 / 10.0)
         .unwrap_or(12.0);
 

src/tables/detect_heuristic.rs

@@ -654,9 +654,12 @@ fn has_consistent_columns(cells: &[Vec<String>]) -> bool {
         *count_freq.entry(count).or_insert(0) += 1;
     }
 
+    // Break ties by preferring higher column count for deterministic output
     let most_common_count = count_freq
         .iter()
-        .max_by_key(|(_, freq)| *freq)
+        .max_by(|(count_a, freq_a), (count_b, freq_b)| {
+            freq_a.cmp(freq_b).then_with(|| count_a.cmp(count_b))
+        })
         .map(|(count, _)| *count)
         .unwrap_or(0);
 

src/tables/detect_rects.rs

@@ -88,10 +88,13 @@ pub(crate) fn cluster_rects(
         groups.entry(uf.find(i)).or_default().push(i);
     }
 
-    groups
-        .into_values()
-        .filter(|g| g.len() >= min_size)
-        .collect()
+    // Sort by root index for deterministic output order
+    let mut result: Vec<(usize, Vec<usize>)> = groups
+        .into_iter()
+        .filter(|(_, g)| g.len() >= min_size)
+        .collect();
+    result.sort_by_key(|(root, _)| *root);
+    result.into_iter().map(|(_, g)| g).collect()
 }
 
 /// Detect tables from explicit rectangle (`re`) operators in the PDF.