I started a similar effort some time ago here:
I didn’t include the Drovandi et al. (2011) method, because I thought for some reason that the similar method by Del Moral et al. (2011) was somewhat superior. Anyway, I usually find kernel-based ABC methods (also included in the package) to be much more sample efficient than MCMC-ABC methods. @francesco.alemanno would you be interested in joining forces to write a really nice and feature-rich ABC package? I like collaborative projects 
.