Remove spurious collision-related code after ultrathink review

After 2 weeks of collision/stale-data fixes, reviewed all changes to identify
spurious code that is no longer needed after content-hash resolution was implemented.

**Removed:**
1. countReferences() function from collision.go (lines 274-328)
   - Was used for reference-count based collision scoring
   - Completely unused after switching to content-hash based resolution (commit 2e87329)
   - Still exists in duplicates.go for deduplication (different use case)

2. ReferenceScore field from CollisionDetail struct
   - Marked as DEPRECATED but never removed
   - No longer used by ScoreCollisions() which now uses content hashing

3. TestCountReferences and TestCountReferencesWordBoundary tests
   - Tested the now-deleted countReferences() function
   - No longer relevant

**Fixed:**
- Updated CheckpointWAL comments to remove misleading "staleness detection" claim
  - Staleness detection uses metadata (last_import_time), NOT file mtime
  - CheckpointWAL is still valuable for data persistence and WAL size reduction
  - Comments now accurately reflect actual benefits

**Verified:**
- All tests pass (internal/storage/sqlite)
- Content-hash collision resolution still works correctly
- No behavioral changes, just cleanup

🤖 Generated with [Claude Code](https://claude.com/claude-code)

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

internal/storage/sqlite/collision.go

@@ -31,7 +31,6 @@ type CollisionDetail struct {
 	IncomingIssue     *types.Issue  // The issue from the import file
 	ExistingIssue     *types.Issue  // The issue currently in the database
 	ConflictingFields []string      // List of field names that differ
-	ReferenceScore    int           // Number of references to this issue (for scoring) - DEPRECATED
 	RemapIncoming     bool          // If true, remap incoming; if false, remap existing
 }
 
@@ -271,62 +270,6 @@ func ScoreCollisions(ctx context.Context, s *SQLiteStorage, collisions []*Collis
 	return nil
 }
 
-// countReferences counts how many times an issue ID is referenced
-// Returns: text mentions + dependency references
-func countReferences(issueID string, allIssues []*types.Issue, allDeps map[string][]*types.Dependency) (int, error) {
-	count := 0
-
-	// Count text mentions in all issues' text fields
-	// Use word boundary regex to match exact IDs (e.g., "bd-10" but not "bd-100")
-	pattern := fmt.Sprintf(`\b%s\b`, regexp.QuoteMeta(issueID))
-	re, err := regexp.Compile(pattern)
-	if err != nil {
-		return 0, fmt.Errorf("failed to compile regex for %s: %w", issueID, err)
-	}
-
-	for _, issue := range allIssues {
-		// Skip counting references in the issue itself
-		if issue.ID == issueID {
-			continue
-		}
-
-		// Count mentions in description
-		count += len(re.FindAllString(issue.Description, -1))
-
-		// Count mentions in design
-		count += len(re.FindAllString(issue.Design, -1))
-
-		// Count mentions in notes
-		count += len(re.FindAllString(issue.Notes, -1))
-
-		// Count mentions in acceptance criteria
-		count += len(re.FindAllString(issue.AcceptanceCriteria, -1))
-	}
-
-	// Count dependency references
-	// An issue can be referenced as either IssueID or DependsOnID
-	for _, deps := range allDeps {
-		for _, dep := range deps {
-			// Skip self-references
-			if dep.IssueID == issueID && dep.DependsOnID == issueID {
-				continue
-			}
-
-			// Count if this issue is the source (IssueID)
-			if dep.IssueID == issueID {
-				count++
-			}
-
-			// Count if this issue is the target (DependsOnID)
-			if dep.DependsOnID == issueID {
-				count++
-			}
-		}
-	}
-
-	return count, nil
-}
-
 // deduplicateIncomingIssues removes content-duplicate issues within the incoming batch
 // Returns deduplicated slice, keeping the first issue ID (lexicographically) for each unique content
 func deduplicateIncomingIssues(issues []*types.Issue) []*types.Issue {