Fix coalescing merge tree segfault for large strings

[scanhex12]

Changelog category (leave one):

• Critical Bug Fix (crash, data loss, RBAC) or LOGICAL_ERROR

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

Fix coalescing merge tree segfault for large strings. This closes #84582

Documentation entry for user-facing changes

• Documentation is written (mandatory for new features)

[clickhouse-gh]

Workflow [PR], commit [095d70a]

Summary: ❌

job_name	test_name	status	info	comment
Stateless tests (amd_tsan, sequential, 1/2)		failure
	03532_crash_in_aggregation_because_of_lost_exception	FAIL
	Killed by signal (in clickhouse-server.log or clickhouse-server.err.log)	FAIL
Stateless tests (amd_tsan, s3 storage, parallel)		failure
	02443_detach_attach_partition	FAIL
Stateless tests (arm_binary, parallel)		failure
	01825_new_type_json_ghdata_insert_select	FAIL

[antaljanosbenjamin]

• 01825_new_type_json_ghdata_insert_select is flaky
• 02443_detach_attach_partition is flaky

For the OOM kill I have to check the logs.

[antaljanosbenjamin]

As @nikitamikhaylov pointed out, these memory issues are from the CI Logs cluster, not locally.

So memory tracker was already killing things from the very beginning:

2025.08.18 20:32:27.232460 [ 1628 ] {} <Information> Application: Will watch for the process with pid 1634
2025.08.18 20:32:27.233617 [ 1634 ] {} <Information> Application: Forked a child process to watch
2025.08.18 20:32:27.234095 [ 1634 ] {} <Information> StatusFile: Writing pid 1634 to /etc/clickhouse-server/clickhouse-server.pid
2025.08.18 20:32:27.234728 [ 1634 ] {} <Debug> CrashWriter: Sending crash reports is initialized with https://crash.clickhouse.com/ endpoint (anonymized)
2025.08.18 20:32:27.234913 [ 1634 ] {} <Debug> Application: Sending logical errors is enabled
2025.08.18 20:32:27.737934 [ 1634 ] {} <Information> Application: Starting ClickHouse 25.8.1.1 (revision: 54501, git hash: a06ab022194df35e98ef285cabcbfe4b444103c3, build id: 0559120BED547127CDC20CB27639C1C036DA93FF), PID 1634
...
2025.08.18 20:32:49.825299 [ 2204 ] {BgDistSchPool::0cb96e5d-0265-4976-be16-20bc42b2bebb} <Error> system.opentelemetry_span_log_sender.DistributedInsertQueue.default: Code: 241. DB::Exception: Received from kng4alm55c.us-east-2.aws.clickhouse-staging.com:9440. DB::Exception: User memory limit exceeded: would use 29.84 GiB (attempt to allocate chunk of 1.19 MiB bytes), maximum: 29.80 GiB. OvercommitTracker decision: Memory overcommit has not freed enough memory. Stack trace:

Therefore I think the issue is not connected to the fix, but I will restart the job to be more confident.

The new test I think should have run in that batch, but it didn't run yet. If it fails again, we might have a problem.