Task: Assemble the package once every input is present

prepare_ore_package downloads whatever it is handed and packs it. If a producer has not finished, or a producer failed, the step cannot tell: it either reads a missing key and throws, or it packs what is there and the engine discovers the gap when it reads the file. A run that is still gathering and a run that is broken look the same until the engine speaks.

Make the step consume the run's declared artifact set, run only when every artifact exists, and name the ones that do not.

• The step dispatches only when every artifact the run declared exists.
• When one is missing, the run reports which artifact is missing and which step produces it, rather than failing inside the engine.
• The package is assembled from storage, so the step holds no state of its own that a restart would lose.
• The engine's working directory is still the layout ORE reads: the run document and its configuration under inputPath, and the run's own data at the paths the run document names.

This is the task that makes the waiting visible, and it depends on both foundations: the declaration says what the run needs, and the graph says whether each producer has finished. Assembling from storage rather than from a staging directory built inside one call is what lets the producers run anywhere and in any order.

ores.ore.store::declared_data_files already reads the names the run document gives the run's own data, and archive_layout already places the configuration documents where the engine reads them. The pieces the earlier change built are the ones this step keeps; what changes is where the bytes come from and when the step is allowed to start.

Every number this task claims about a tree, with the command that produced it and the commit it was measured at. A number without both is not a measurement, because a clean-up record describes a tree and trunk moves under a branch: a consumer count, a gate's component count and a suite's size all change with no code changing, and the reader of a bare number cannot tell which tree it came from. Where the raw output matters, put it in an evidence/ file beside the task and name that file in the Command column.

Manual QA scenarios (scaffolded via compass add test_scenario) that verify this task. Link new ones here as they're created; the scenario doc itself links back via its "Verifies task" field.