feature request - fault tolerant builds for loop_tasks when the files all exist

No immediate priority on this because I think we all have a way around this (e.g., deleting all of the files/targets).

Sometimes the tasks we are using `loop_tasks` for are brittle (e.g., downloads on shaky services). If a rebuild is triggered because some params or code changes and you already have files for the tasks, the fault tolerant part of `loop_tasks` is skipped over and you go right to `scmake`, which will only try once and error on the first fail. Ideally I think we'd still get the fault tolerant part of this function if the targets already exist but they are stale.

Start by tracing the `loop_tasks` implementation and the point where it hands existing or stale targets to `scmake`. Reproduce a rebuild with all task files present and an intermittent download failure, then verify that the fault-tolerant retry behavior is still used and the build completes when a retry succeeds.

Written by the indexing model from the issue text.