[RFC 0005] API for Continued Sampling from R and Python

[andrewherren]

• R continue_sampling.bartmodel / .bcfmodel with num_gfr / num_burnin / num_mcmc + per-forest/general params lists matching bart() / bcf().
• Python BARTModel.continue_sampling / BCFModel.continue_sampling with the same signature shape as BARTModel.sample / BCFModel.sample.
• Runtime guard: recover structural params fixed by the model (e.g. num_trees) from the model; warn + ignore attempts to change them.
• Allow num_gfr > 0 on continuation (XBART/XBCF and re-warm-start use cases).
• Sample accumulation: append to BART/BCFSamples; update internal num_gfr / num_burnin / num_mcmc counts.
• Document deserialized-continuation caveat: many hyperparameters not explicitly reset by parameter lists will fall back to package defaults.