[8.x] Fix the default datasource setting in the docs, explain startup usage

[mkurz]

• Backport of Fix the default datasource setting in the docs, explain startup usage #910 to 8.x.

The docs named a non-existent ebeanconfig.datasource.default setting to choose the default Ebean server, but it's play.ebean.defaultDatasource (as reported in #145).

Also add a section about using Ebean during application startup (see #457 and #65):

• Components that use Ebean while they get created, e.g. eagerly bound ones or ones created via Guice's static injection (which runs even before eager singletons get created), have to depend on DynamicEvolutions, which creates the Ebean databases. Otherwise Ebean tries to create the database itself from its own configuration, which usually fails and leaves Ebean unusable until the JVM restarts. Later NoClassDefFoundError: Could not initialize class ... errors only point back to that first failure.
• That doesn't mean that the evolutions have been applied. In dev mode, Play doesn't wait for pending evolutions (not even for components depending on ApplicationEvolutions), and can't show the evolutions page if the application fails to start. So the docs recommend not to use the database schema while components get created, or to let Play apply evolutions automatically, limited to dev mode via PlayKeys.devSettings, which is also what Evolutions not run property on play 2.8 #220 asks for.

Testing

• Reproduced on current main: a repository created via Guice's requestStaticInjection that calls DB.getDefault() in its constructor makes the application fail to start, in prod as well as in dev mode. With a dependency on DynamicEvolutions, as in the new docs sample, it starts.
• Reproduced that in dev mode with pending evolutions, a component querying the schema while it gets created makes the application fail to start, also when depending on ApplicationEvolutions. With autoApply, it starts.
• The new docs sample gets compiled by the docs build; cd docs && sbt "validateCode; test" passes.

[mergify]

This pull request does not currently match the merge queue conditions, so it cannot be queued from here. The box comes back if it matches again.