Allow __splink__df_concat to be computed without linker

[RobinL]

We plan to allow the user to do some forms of exploratory analysis without needing to create a linker (like profile_columns and various types of blocking analysis e.g. #2136 )

But this means that __splink__df_concat needs to be computed without the linker.

At the moment, this requires a lot of code that's confusing to read and will be repetitive:

splink/splink/profile_data.py

Lines 237 to 267 in 66ec54f

	tables = ensure_is_list(table_or_tables)
	tables = db_api.process_input_tables(tables)
	splink_df_dict = db_api.register_multiple_tables(tables)
	input_dataframes = list(splink_df_dict.values())
	input_aliases = list(splink_df_dict.keys())
	input_columns = input_dataframes[0].columns_escaped
	if not column_expressions:
	column_expressions_raw = input_columns
	else:
	column_expressions_raw = ensure_is_list(column_expressions)
	column_expressions = expressions_to_sql(column_expressions_raw)
	pipeline = CTEPipeline(input_dataframes)
	cols_to_select = ", ".join(input_columns)
	template = """
	select {cols_to_select}
	from {table_name}
	"""
	sql_df_concat = " UNION ALL".join(
	[
	template.format(cols_to_select=cols_to_select, table_name=table_name)
	for table_name in input_aliases
	]
	)
	pipeline.enqueue_sql(sql_df_concat, "__splink__df_concat")

Issue can be addressed by removing the need for a linker to compute __splink__df_concat, giving us reusable code that can be used for profile_columns, blocking analysis etc.

[RobinL]

We want vertically_concatenate_sql to be modified so it doesn't take a linker as an argument, but the functions like vertically_concatenate.compute_df_concat shuld still take the linker as an argument