Hi and welcome! Being able to use Julia in PDAL pipelines would be great. I thought about doing that as well a while ago, but never got around to it.
My understanding is that this is most useful for doing custom steps that are not already available in PDAL. So input and output are probably already covered, and LasIO may not be very helpful, since PDAL’s internal data model is not equal to LAS I believe.
While Python uses a dictionary of Numpy arrays, using a Dict of Arrays in Julia is probably not optimal. I have not looked into PDAL’s data model, but perhaps you can use something similar in Julia? It would be great for performance if PDAL does not have to make any copies.
It is probably also a good idea to ask about the design to the PDAL mailing list, and let them know that you are working on this.