Ability to set chunk size for pdf-to-markdown conversion

[albydeca]

To the best of my knowledge, the LlamaParse constructor's split_by_page parameter will force the parser to output either one Document per page (if the value is true) or one Document for the entire pdf (if the value is false).

This may be too coarse-grained for some applications. It would be great to allow the user to set a chunk size and a chunking strategy similar to what Unstructured are doing.

This could help especially when the pages to be extracted are dense, chunks are very large, and the embedding vector becomes too generic.