 * Move objects specific to examples into appropriate namespaces

 * Use Rcpp attributes, which also allows for removal of rngR.h

 * Make the data a data component of the package

 * Add log evidence estimation capabilities and add additional integration
   methods to IntegratePathSampling

 * Add more examples

 * Make it easier to add additional parameters (for example the covariance
   matrix used in a random walk proposal in MCMC) and to adapt these parameters
   using the population of particles

 * For blockpfGaussianOpt et al, we could be fancy and return an S3 object
   with a plot method.  Maybe later.

 * Add support for an arma::mat version of the particle values.

 * Add other interesting SMC algorithms, e.g.
   * The iterated auxiliary particle filter of Guarniero et al. (2015)
   * Divide-and-Conquer SMC (Lindsten et al., 2016)
   * Particle MCMC (Andrieu et al., 2010)