Better support for default schemas

[stevennic]

We should be able to specify a different default schema at the DB level from the beginning, and have all operations operate in that scope without further annotation. If I want to set a default schema on initialization with Postgres, I have to use the options binding to specify -c search_path=<schema> as a string. There should be a native way to set the schema.

There are also hard-coded assumptions that we are in the default schema. For example, dbapiprovider.py:split_table_name starts with this:

    def split_table_name(provider, table_name):
        if isinstance(table_name, str): return provider.default_schema_name, table_name

Given a table name, it hard codes its schema scope to the default_schema_name, even if the current DB connection is using a different schema. This is called by DBAPIProvider.table_exists to check if a table should be created when mappings are being generated. The default schema hard-code causes this to return False, as it's looking for the tables in the wrong schema, causing it to attempt to create a table that already exists (because the connection it uses to attempt to then create it is in the correct schema).

As a workaround, I subclassed the PGProvider and overrode split_table_name to return the current schema, but the correct solution is for the native Provider to handle this.

class MyPGProvider(PGProvider):
    def __init__(self, *args, schema: str = None, **kwargs):
        self._schema = schema
        super().__init__(*args, **kwargs)

    def split_table_name(self, table_name):
        if self._schema:
            return self._schema, table_name
        else:
            return super(self.__class__.__base__, self).split_table_name(table_name)  # Let it use the default schema