Michał Jadwiszczak

Changelog: - get_all_referenced_user_types() returns set of `user_type`s instead of names - test: recreating all UDTs in a new keyspace instead of dropping test keyspace and recreating it - test: indexing...

> #### Failed Tests (2/40220): > * [topology_experimental_raft.test_tablets.debug.4](https://jenkins.scylladb.com//job/scylla-master/job/scylla-ci/8582/testReport/junit/%28root%29/non-boost%20tests/Tests___Unit_Tests___topology_experimental_raft_test_tablets_debug_4) [🔍](https://github.com/scylladb/scylladb/issues?q=is:issue+is:open+topology_experimental_raft.test_tablets.debug.4) > * [test_tablet_split](https://jenkins.scylladb.com//job/scylla-master/job/scylla-ci/8582/testReport/junit/%28root%29/test_tablets/Tests___Unit_Tests___test_tablet_split) [🔍](https://github.com/scylladb/scylladb/issues?q=is:issue+is:open+test_tablet_split) Caused by https://github.com/scylladb/scylladb/issues/18085

Chanelog: - extracted topological sorting algorithm to generic implementation - all required changes to use generic topological sorting in cql_type_parser (futurized build() function and adopted it to use sorting) -...

Changelog: - fixed a bug in cql_type_parser - added a test for topological sorting - applied smaller comments like genlty cleaning `ref_count_map` in sorting, etc... - (cc @michoecho @piodul) `tools::load_schemas()`...

Changelog: Addressed @piodul comments

Changelog: - now cdc log tables cannot be described directly with `DESC TABLE some_table_scylla_cdc_log` - description of base cdc table contains `ALTER TABLE some_table_scylla_cdc_log WITH ...` including all schema's parameters...

Changelog: Applied requested changes, mostly comments improvement, extended error message when trying to describe cdc log table

> It's also worth mentioning that this procedure requires proper support in cqlsh. The current implementation intercepts both `DESC SCHEMA` and `DESC SCHEMA WITH INTERNALS` and handles them in a...

> Better to harden the create function so it returns a documented exception if it fails to create a new scheduling group. Maybe it's not the best place to discuss...

Closing the PR without merging. We are not going to add a test to OSS.