[Task manager] Evaluate the impact of Ephemeral Tasks

[gmmorris]

The Ephemeral Task feature was released in #90888 (PR: #97311) under a feature flag and is disabled by default.

In order to open this feature up we need to verify its performance and sustainability at scale.

Acceptance Criteria

1. Before this is tested, we need to define a concrete experiment which we should write up as an RFC and share with tech leads for review. Such an experiment needs a clear hypothesis, a reproducible test plan and defined metrics for success.
2. Once an experiment is defined, we should run it with both the feature enabled, and disabled, for comparison.
3. After the experiment the results will be shared with the team, so a decision can be made in regards to enabling the feature by default.

Possible Outcomes

The way I see it (open to debate, obviously) we have three possible outcomes:

1. We decide the feature is scalable, sustainable and valuable - enabling it by default.
2. We decide the feature is scalable and sustainable, but not valuable enough - we delete it (yay, less code to maintain, and we've learned something new)
3. We decide the feature is valuable, but isn't scalable or sustainable enough (due to added complexity, for instance) - further discussion is needed as to whether the value is sufficient to warrant further technical work that might be needed to address the lack of scalability/sustainability.

[elasticmachine]

Pinging @elastic/kibana-alerting-services (Team:Alerting Services)