Fix keep_size for arviz structures. (#4006)

* Fix posterior pred. sampling keep_size w/ arviz input.

Previously posterior predictive sampling functions did not properly
handle the `keep_size` keyword argument when getting an xarray Dataset
as parameter.

Also extended these functions to accept InferenceData object as input.

* Reformatting.

* Check type errors.

Make errors consistent across sample_posterior_predictive and fast_sample_posterior_predictive, and add 2 tests.

* Add changelog entry.

Co-authored-by: Robert P. Goldman <rpgoldman@sift.net>

RELEASE-NOTES.md

@@ -11,6 +11,7 @@
 ### New features
 - Introduce optional arguments to `pm.sample`: `mp_ctx` to control how the processes for parallel sampling are started, and `pickle_backend` to specify which library is used to pickle models in parallel sampling when the multiprocessing cnotext is not of type `fork`. (see [#3991](https://github.com/pymc-devs/pymc3/pull/3991))
 - Add sampler stats `process_time_diff`, `perf_counter_diff` and `perf_counter_start`, that record wall and CPU times for each NUTS and HMC sample (see [ #3986](https://github.com/pymc-devs/pymc3/pull/3986)).
+- Extend `keep_size` argument handling for `sample_posterior_predictive` and `fast_sample_posterior_predictive`, to work on arviz InferenceData and xarray Dataset input values. (see [PR #4006](https://github.com/pymc-devs/pymc3/pull/4006) and [Issue #4004](https://github.com/pymc-devs/pymc3/issues/4004).
 
 ## PyMC3 3.9.2 (24 June 2020)
 ### Maintenance