idea for new design of blob v2's input and output schema

This seems fine as input. However, I think the weirdest thing about blobs is going to be the fact that the output is not the same as the input. Right now the input is:

{
  "data": large_binary,
  "uri": utf8,
}

However, the output is the descriptions and the descriptions contain a number of fields (e.g. blob_id) which are internal details that don't make sense to the user. I'm wondering if we can, at some point, unify these two things. For example, when the user reads blob data, we can convert the blob_id, data file, etc. into a URI. So what if we add to the blob array position and length (with length=-1 meaning the whole file) and so the input and output are the same? The only difference is that the input might be a mix of data and uri but the output would always have data be null and uri be set?

Originally posted by @westonpace in https://github.com/lance-format/lance/pull/5491#pullrequestreview-3592810872

Start by reviewing the blob v2 input and output schema described in the issue, including the data, uri, blob_id, array position, and length fields. Compare the proposed unified representation with the current blob descriptions; done would require an agreed schema design that specifies how input and output should represent blob data.