With datasets stored by uuid, a dataset's uuid is where its file lives.
A job could set its output's uuid through tool-provided metadata
(galaxy.json), and Galaxy applied it when finishing the job and in the
extended metadata script. An upload could likewise request any uuid,
including one already in use.
Set a requested upload uuid when Galaxy creates the dataset instead,
after checking that it is a valid uuid no dataset has yet, and stop
applying a uuid from tool-provided metadata. The upload tool no longer
passes the requested uuid back through galaxy.json, so the library path
upload sets it before creating the dataset as the other uploads do.
Nuwan Goonasekera
b792c7297e