Adds ranges for PPTSS sampling

[jzonthemtn]

Adds ranges for PPTSS sampling for #103.

[wrigleyDan]

Changes look good, the code works, all my testing was successful.

As I went through the changes and noticed the renaming of the classes I got confused (mainly because of the QueriesQueryQuery part in the class AllQueriesQueryQuerySamplerParameters - is that intentional or did one Query too many slip into the name?). That lead to the following thought: I think three sampling methods is what we ultimately want: a random sampler, a frequency weighted sampler (what PPTSS is), a "Top N" sampler picking just the most frequent N queries. This is also described in Stavros' document about the eyeballing tool.

How could that look like:

• The AllQueriesQuerySampler becomes the RandomQuerySampler that picks N queries randomly out of all queries.
• A new TopNQuerySampler is introduced that picks the N most frequent queries out of all queries.

When doing so I think renaming some of the classes might add clarity. For example:

• AllQueriesQueryQuerySamplerParameters --> RandomQuerySamplerParameters
• ProbabilityProportionalToSizeParametersQuery --> ProbabilityProportionalToSizeSamplerParameters

These changes should happen in a separate branch. I wanted to make this suggestion here to get the discussion started. What do you think?

[jzonthemtn]

AllQueriesQueryQuerySamplerParameters

I fixed the name of this class to take out one of the Query names and pushed it up.

I agree about the samplers and wrote:

• Add a RandomQuerySampler query sampler #113 Add a RandomQuerySampler
• Add a TopNQuerySampler query sampler #114 Add a TopNQuerySampler

I want to leave the AllQueriesQuerySampler sampler in primarily just for debugging. I know it's probably not very useful for search relevance engineering, but I have found it useful for testing.

So if it's ok with you, we can merge this PR and I'll add the other samplers in other pull requests.

[wrigleyDan]

LGTM 👍