Separate auto-configuration of script-based DataSource initialization from the auto-configuration of the DataSource

[wilkinsona]

Currently, the auto-configuration of script-based DataSource initialization is intermingled with the auto-configuration of the DataSource itself. This means that the spring.datasource.* properties are a mixture of those that are used to create the DataSource and those that are then used to initialise it using DDL and DML scripts. We should separate out the auto-configuration and the properties that are specific to our script-based initialization.

In addition to making it clearer which properties are for initialization and which are for general data source configuration, this will align the structure of the auto-configuration for our script-based initialization with how things are structured for Flyway and Liquibase.

[wilkinsona]

This issue is noteworthy as all of the spring.datasource properties that are specific to initialization are being deprecated with spring.sql.init replacements. The separate usernames and passwords for schema and data scripts are not being replaced. Instead, the same credentials, configured using spring.sql.init.username and spring.sql.init.password, are used for DDL (schema) and DML (data) scripts. If users need separate credentials for DDL and DML, they can define their own, appropriately configured data source initializer beans.

[wilkinsona]

Re-opening to include SqlInitializationAutoConfiguration in test slices that use JDBC.