Munge functionality for the particle filter.

[Averydx]

This PR adds munge functionality for PipelineOutput.

• Added dataframe property to PipelineOutput
• Added classes to support RealizationStrategy
• Added classes to support ParameterStrategy
• Added new munge function munge_pipeline_output

RealizationSelector and ParameterSelector are properties of PipelineOutput whereas
the other selectors are still properties of the rume.

Example usage

``` realization` = particle_filter_output.select.all()
time = rume.time_frame.select.all().group("day").agg()
geo = rume.scope.select.all()
quantity = particle_filter_output.param_select.by_name("log_beta")

df = munge_pipeline_output(particle_filter_output,realization,geo,time,quantity) ```

[Averydx]

I added a munge function to operate on both PipelineOutput and Output. It just adds a realization dimension if the result is an Output and does the same processing. Unfortunately it is slower than standard munge.

The average runtime of munge_combined was 30.15 seconds vs 23.5 seconds for the old munge.