Support for a JDBC source

When defining a DBT model we can use the source macro to reference an existing table in Clickhouse using database and table name, but in Clickhouse a NamedDataSource can be defined an a table can be referenced using not only `.` but also `jdbc('', '')`.

As it is now in the dbt-clickhouse adapter, there is no way (afaik) to keep the concept that this jdbc bridge reference is a source and access the data, because there is now source macro call we can write that would make it work.

There is a number of solution we could imagine and we'd need to talk about what would be best, my main option would be to :
* use the loader tag in the sources yaml definition, to change the source resolution in the case where loader=jdbc (it is accessible in the SourceDefinition object even if documentation states that it's only for doc purposes, I guess we could use it) and override the ClickHouseRelation.render method.

but we could also
* define an additional macro like jdbc_source (postgresql_source, mysql_source as well I guess) that would proxy the source resolution

I don't think this is doable apart from within the adapter, as it would need to override the rendering process of the relation, and I do think that abandoning the concept of source for jdbc/postgresql/mysql... engines would be a bit sad.

Start by tracing the dbt source macro and the adapter's ClickHouseRelation.render path, then inspect how SourceDefinition.loader is exposed for sources defined in YAML. The issue needs a decided design before implementation; done should mean a JDBC-backed source can resolve its named data source and table or query while retaining source semantics.

Written by the indexing model from the issue text.