Improve handling of sparse tables

Improve robustness when tables contain empty cells.

---

## ✅ Challenge Spec

## 🧩 Problem
Describe the limitation / failure mode this issue addresses (include real-world context where possible).

## 🎯 Goal
State the expected improvement clearly (accuracy, stability, correctness, etc.).

## ✅ Acceptance Criteria
- Improvement demonstrated on **at least 2** representative sample PDFs (synthetic PDFs acceptable if shareable)
- Add/update **unit or integration tests** for the behaviour
- No regression in existing test suite / CI checks
- Update docs/Javadoc if behaviour or configuration changes

## 🧪 Validation
- Provide before/after output comparison (or screenshot/log excerpt)
- Link test evidence and sample inputs (or reproducible steps)

---

---

**Challenge Note:** This issue is part of the ExtractPDF4J Global Build Challenge 2026.
To participate: fork the repo → create a branch → submit a PR referencing this issue → Star the repo.