Return TEXT instead of VARCHAR columns

In DuckDB the canonical name for an unlimited size text column is
`VARCHAR`[1], in Postgres this is `TEXT`[2]. In DuckDB `TEXT` is simply
an alias for `VARCHAR` type, and there's no way to know what was
provided by the user. In Postgres these types are actually distinct,
although behave exactly the same for unlimited length. Basically
everyone uses `TEXT` instead of `VARCHAR`.

Currently we convert the DuckDB type to a Postgres `VARCHAR`. In many
cases this doesn't really matter, because pretty much all clients handle
VARCHAR and TEXT the same too. There's one place where this leaks
through though: DDL coming from a query. For example if you do a CTAS
with a DuckDB query the resulting table columns will be of type
`character varying` instead of `text`[3].

[1]: https://duckdb.org/docs/sql/data_types/text.html
[2]: https://www.postgresql.org/docs/current/datatype-character.html
[3]: #556 (comment)

src/pgduckdb_types.cpp

@@ -1015,7 +1015,7 @@ GetPostgresArrayDuckDBType(const duckdb::LogicalType &type) {
 	case duckdb::LogicalTypeId::UINTEGER:
 		return INT8ARRAYOID;
 	case duckdb::LogicalTypeId::VARCHAR:
-		return type.IsJSONType() ? JSONARRAYOID : VARCHARARRAYOID;
+		return type.IsJSONType() ? JSONARRAYOID : TEXTARRAYOID;
 	case duckdb::LogicalTypeId::DATE:
 		return DATEARRAYOID;
 	case duckdb::LogicalTypeId::TIMESTAMP:
@@ -1064,7 +1064,7 @@ GetPostgresDuckDBType(const duckdb::LogicalType &type) {
 	case duckdb::LogicalTypeId::UINTEGER:
 		return INT8OID;
 	case duckdb::LogicalTypeId::VARCHAR:
-		return type.IsJSONType() ? JSONOID : VARCHAROID;
+		return type.IsJSONType() ? JSONOID : TEXTOID;
 	case duckdb::LogicalTypeId::DATE:
 		return DATEOID;
 	case duckdb::LogicalTypeId::TIMESTAMP:

test/regression/expected/materialized_view.out

@@ -1,4 +1,4 @@
-CREATE TABLE t(a INT, b VARCHAR);
+CREATE TABLE t(a INT, b TEXT);
 INSERT INTO t SELECT g % 100, MD5(g::VARCHAR) FROM generate_series(1,1000) g;
 SELECT COUNT(*) FROM t WHERE a % 10 = 0;
  count 

test/regression/sql/materialized_view.sql

@@ -1,4 +1,4 @@
-CREATE TABLE t(a INT, b VARCHAR);
+CREATE TABLE t(a INT, b TEXT);
 INSERT INTO t SELECT g % 100, MD5(g::VARCHAR) FROM generate_series(1,1000) g;
 SELECT COUNT(*) FROM t WHERE a % 10 = 0;
 CREATE MATERIALIZED VIEW tv AS SELECT * FROM t WHERE a % 10 = 0;