The most common use case is when people want to configure replication and sharding. Currently they have to use index templates for that.
One quick and easy-to-implement solution would be to add a configuration option:
Later, we can introduce more powerful solutions such as HSEARCH-3306.
How should we interpret the file path: relative to the classpath root? Relative to the working directory? As an URL?
Should we take the settings into account when validating the schema?