Adding CTFipsQuery class and demo

[Niklewa]

In this PR, I added CTFipsQuery, a demo that exemplifies its use, and appropriate tests.

One challenge with CTFipsQuery is that it seems rather slow, the process of creating a comparison takes around 30-60 seconds.

[rfl-urbaniak]

Please expand on the profiling, along the lines of what's in grouped_predictions.ipynb, along the lines of:

def basic_run():
    ci = CausalInsight(
        outcome_dataset=outcome,
        intervention_dataset=intervention,
        num_samples=num_samples,
        smoke_test=smoke_test,
    )

    ci.get_tau_samples()
    percent_calc = ci.slider_values_to_interventions(intervened_percent=50, year = year)
    assert percent_calc['intervened_transformed'] > -1
    
    for fips in texas:
        ci.get_fips_predictions(intervened_value=percent_calc['intervened_transformed'],
                                 fips=fips, intervention_is_percentile=True)


profiler_basic = cProfile.Profile()

profiler_basic.enable()
basic_run()
profiler_basic.disable()

profiler_basic.print_stats(sort='cumulative')

[Niklewa]

Apart from adding a more thorough time performance test for CTFipsQuery, I also included an additional test for DataGrabber to make it easier to detect mismatches in data size between variables at the same level. Additionally, I fixed the problem with data mismatch (which explains the number of changes introduced by this PR) by modifying the VariableCleaner class.

[rfl-urbaniak]

So the main reason of slowdown is multiple use of reduce:

4.174 0.035 4.174 0.035 {method 'reduce' of 'numpy.ufunc' objects}

at some point downstream we might investigate further if this is a major issue.