corrects error with unpack_results argument

[samland1116]

I was able to get the unpack_results functionality to work as intended (prior to this update) like:

# Define a bootstrapper with custom parameters.
boot = teehr.Bootstrappers.Gumboot(
    reps=500,
    quantiles=[0.05, 0.5, 0.95],
)

metrics_sdf = ev.metrics.query(
    group_by=[
        "primary_location_id",
        "configuration_name"
    ],
    include_metrics=[
        teehr.DeterministicMetrics.RelativeBias(
            bootstrap=boot,
            unpack_results=True # defined here instead
        )
    ]
).to_sdf()

When applying the bootstrapper in teehr.querying.metric_format.apply_aggregation_metrics(), we were not checking if the bootstrap.unpack_results attribute was set or not. The unpacking hinged solely on if you defined the argument when instantiating the metric function you are bootstrapping.

I updated teehr.querying.metric_format.apply_aggregation_metrics() to check the bootstrap attributes for unpack_results. This allows you to define the unpack_results arg when instantiating the bootstrap and/or when defining the metric you apply the bootstrap to (i.e. the code Matt attached in the issue works in addition to the above).

[samlamont]

I think the BootstrapBasemodel may be at fault here, in that it inherits the MetricsBasemodel which has the unpack_results field. So in the above example we have boot model that is added to the RelativeBias metric, but they have redundant fields since they both inherit the MetricsBasemodel.

Seems like maybe we should remove them from the BootstrapBasemodel (or MetricsBasemodel?) so we only have to check the condition once. Maybe we can discuss next week