There are a handful of examples in the ILM documentation that could result in
rolling over indices more quickly than we might normally recommend,
contributing to over-sharding in cases where the examples are copied without
modification. This change makes some numbers bigger to try and avoid this.
This adds an example to drive home the semantics of `min_age` and explain
how actions from one phase must complete before `min_age` is even tested
closes#34020.
