Sizing the filesystem-writing gap instead of starting it

writing the filesystem was next on the backlog, and it turned out to be the wrong thing to start. examples/superblock.rs asks how much of a superblock a writer would have to produce that nothing here can name.

Three real packages, identical answers: 38 of 1024 bytes carry a cited field; 95 further bytes are non-zero and unaccounted for, in sixteen runs. Two of the runs are thirty-two bytes wide and sit where a digest would - a shape, not a meaning.

So this is the reader-only-source problem, not effort remaining, and the backlog now says so. The container spent months in this exact position and was unblocked in an afternoon when an open-source writer turned up; a reader walks a filesystem by following four numbers and never looks at the other 95 bytes, so no number of readers will supply them.

The example stays because the question recurs. It answers one, it does not derive a format: nothing from it went into data/.