Ensure outcome metrics properly override

[mikewilli]

We had a problem where jetstream was reporting errors with urlbar_impressions not having a statistic defined despite a statistic configured in the outcome definition, affecting only experiments with the Firefox Suggest outcome. This occurred following the addition of a top-level metric with the same name, which appeared to affect the outcome metric's definition, which I believe should have overridden the top-level definition.

The metric in definitions/firefox_desktop.toml

[metrics.urlbar_impressions]
select_expression = "COALESCE(COUNTIF(is_terminal), 0)"
data_source = "urlbar_events"
friendly_name = "URL Bar Impressions"

and in jetstream/outcomes/firefox_desktop/firefox_suggest.toml

[metrics.urlbar_impressions]
select_expression = "COUNTIF(is_terminal)"
data_source = "urlbar_events"
description = "The number of times a user exits the urlbar dropdown menu, either by abandoning the urlbar, engaging with a urlbar result, or selecting an annoyance signal that closes the urlbar dropdown menu"
friendly_name = "Urlbar search sessions"
exposure_basis = ["exposures", "enrollments"]
statistics = { deciles = {}, bootstrap_mean = {} }

We should ensure that metric definitions in outcomes override top-level metric definitions as expected.

[mikewilli]

Also of note, there are population ratio metrics in the outcome definition that reference the urlbar_impressions metric, and it could be that these are the ones that are only referencing the upstream metric for some reason.