Compare data generated with HybridChunker versus our existing custom chunking implementation

[bbrowning]

We need to compare the data samples generated with Docling's new HybridChunker versus our existing Docling / custom chunking implementation for the same set of input document(s).

Do the new chunks look reasonable? Are the chunks substantially larger or smaller than before? This may be a manual spot-checking effort, unless we come up with a reasonable effort integration with an external tool to analyze our generated data samples.