move index creation for word table to tokenizer

This introduces a finalization routing for the tokenizer
where it can post-process the import if necessary.

lib-sql/tokenizer/legacy_tokenizer_indices.sql

@@ -0,0 +1,2 @@
+CREATE INDEX {{sql.if_index_not_exists}} idx_word_word_id
+  ON word USING BTREE (word_id) {{db.tablespace.search_index}};