panic in the face of type cycles

[ahl]

Closes #150

These can arise because the schemars ignores conflicting names and
doesn't use e.g. the mod to qualify the name when necessary. There are
other errors that can result from this such as overwritten types, but
this is the most acute at the moment.

@smklein I don't think this is a great solution, but it may actually be what we want. There an issue in schemars that describes the problem of name conflicts more generally. In this case, however, we actually don't want two types in the schema and #[serde(transparent)] does exactly what we want.

I'm not sure why serde (and schemars) don't do this automatically when they see #[repr(transparent)] but it may be that derive macros don't reliably see that attribute. See also serde-rs/serde#1054

[smklein]

Thanks for fixing - I believe this would have hit my use case, and helped me get to a working solution quickly!