Add support for varying D to fitdistdoublecens

[seabbs]

In most realistic datasets this is likely to vary and so we should support if possible. This would then also better match the functionality in the new (see #67) stan model