do not run wal on remote disks

[CheSema]

Changelog category (leave one):

• Not for changelog (changelog entry is not required)

I have found out here that CH uses wal on S3 disk in tests.

It makes not much sence. It's better to write compact parts instead. Furthermore CH do not use such combination in real work.
In test I see that wal needs to be sync often in order to fulfill the assignment and be robust with restarts.
It is not possible to sync writing in S3. It is possible either finish the file or continue writing in it without fixation of written file part. Interesting fact: rotation works for fixation written file part much better Instead of sync.

I'm going to collect what test is failed and turn them off for s3-disks and other remote disks.

Information about CI checks: https://clickhouse.com/docs/en/development/continuous-integration/

[CheSema]

01130_in_memory_parts_nested
01643_replicated_merge_tree_fsync_smoke
01130_in_memory_parts_partitons
02423_drop_memory_parts
01475_read_subcolumns_storages
01130_in_memory_parts_check
02410_inmemory_wal_cleanup
01508_race_condition_rename_clear_zookeeper_long
01643_merge_tree_fsync_smoke
01130_in_memory_parts_default
01130_in_memory_parts
01600_parts_types_metrics_long

[CheSema]

test_backup_restore_on_cluster
Data race in BackupImpl::writeFile #46040