adjust the README

implement proof of concept for du -bkB

This avoids using cfg with the destination-crate's feature "complete"
in proc-macro "derive".

Quoting cargo:

    using a cfg inside a derive macro will use the cfgs from the
    destination crate and not the ones from the defining crate

In these two instances, the cfg was simply used to avoid emitting dead
code if the "complete" feature is not desired. By passing the "complete"
feature down to the derive crate and evaluating if there, we achieve the
same goal, without having to deal with suprisingly-valued and unexpected
cfgs downstream.

Alternatively, we could just evaluate the flag inside the derive crate,
and pass the feature flag to it.

…en test

derive: make error messages slightly more readable, add manually-driven test

…eflag

remove 'complete' feature flag, make it unconditional

Make 'complete' available to procmacro's output by conversion to module

…ew issues

split 'compile_error' toy into many test cases, demonstrate new issues

clippy: suppress warning from `large_enum_variant`