Fix reading from text index on s3

[CurtizJ]

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):

Fixed direct reading from the text index on object storage.

[clickhouse-gh]

Workflow [PR], commit [6e4e7f6]

Summary: ❌

job_name	test_name	status	info	comment
Stateless tests (amd_binary, ParallelReplicas, s3 storage, parallel)		failure
	03117_analyzer_same_column_name_as_func	FAIL
	Exception in test runner	FAIL
	Killed by signal (in clickhouse-server.log or clickhouse-server.err.log)	FAIL
	Fatal messages (in clickhouse-server.log or clickhouse-server.err.log)	FAIL
Integration tests (amd_binary, 4/5)		failure
	test_merge_tree_s3/test.py::test_merge_canceled_by_s3_errors[node-broken_s3_always_multi_part]	FAIL
Stress test (amd_debug)		failure
	Server died	FAIL
	Hung check failed, possible deadlock found (see hung_check.log)	FAIL
	Killed by signal (in clickhouse-server.log)	FAIL
	Fatal message in clickhouse-server.log (see fatal_messages.txt)	FAIL
	Killed by signal (output files)	FAIL
	Found signal in gdb.log	FAIL
Stress test (amd_ubsan)		failure
	Server died	FAIL
	Hung check failed, possible deadlock found (see hung_check.log)	FAIL
	Killed by signal (in clickhouse-server.log)	FAIL
	Fatal message in clickhouse-server.log (see fatal_messages.txt)	FAIL
	Killed by signal (output files)	FAIL
	Found signal in gdb.log	FAIL
Finish Workflow		failure
	python3 ./ci/jobs/scripts/workflow_hooks/new_tests_check.py	failure

[ahmadov]

LGTM 👍

[CurtizJ]

I've tried to create a minimal repro to add a test. The minimal repro I could create still requires ingestion for about 1 minute. The bug reproduces under certain conditions when stealing of tasks is used in MergeTreePrefetchedReadPool (see here). I tested manually on S3, and this PR fixes Attempt to set read until position before already read data error.