skip empty deduplication tokens

[CheSema]

Due to this injected exception

2026.02.19 23:40:16.383165 [ 8263 ] {95dfe8f2-17fa-4b36-aea9-b31ed884412b} <Error> AsynchronousInsertQueue: Failed parsing for query 'INSERT INTO test_1.async_inserts_native (m) FORMAT Native' with query id 4e00b9b9-6852-4902-a9cc-accb959ce24c. DB::Exception: Query memory tracker: fault injected. Would use 19.29 MiB (attempt to allocate chunk of 4.02 MiB), maximum: 9.31 GiB

A token with 0 rows is added

2026.02.19 23:40:16.392908 [ 8263 ] {95dfe8f2-17fa-4b36-aea9-b31ed884412b} <Test> DedupInfo: setUserToken: token= count=0, instance_id: 4514, async, enabled, level SOURCE,
 rows/tokens 1000/2, in block: null, tokens: 2:[-::1000,-::1000], visited views: 0:[], retried view id: <empty>, original block id: <empty>, unification_stage OLD_SEPARATE
_HASHES

After first token has been filtered we have empty block but 1 token is left.
And this chassert is triggered

if (removed_rows == block.rows())
    {
        chassert(removed_tokens == getCount());

Changelog category (leave one):

• Bug Fix (user-visible misbehavior in an official stable release)

Changelog entry (a user-readable short description of the changes that goes into CHANGELOG.md):

skip empty deduplication tokens

Documentation entry for user-facing changes

• Documentation is written (mandatory for new features)

[clickhouse-gh]

Workflow [PR], commit [a51b298]

Summary: ❌

job_name	test_name	status	info	comment
Stateless tests (arm_asan, azure, parallel)		failure
	00752_low_cardinality_mv_1	FAIL	cidb
	Logical error: Tables UUID does not match after RESTART REPLICA (old: A, new: B) (STID: 4373-3c9f)	FAIL	cidb, issue
Finish Workflow		failure
	python3 ./ci/jobs/scripts/workflow_hooks/new_tests_check.py	failure

[azat]

Should setUserToken be even called for empty chunks?

[CheSema]

Should setUserToken be even called for empty chunks?

I could change it, but code there looks fine. Entry is rolled back, buffer is rolled back, 0 rows are returned.
So I think this patch is just enough.