Translog corruption marker

[vladimirdolzhenko]

Spin-off of #28434:

Today if Lucene index is not good, we mark its store as corrupted by storing a corruption marker. With the presence of the corruption marker, a corrupted copy won't be used as a primary and requires a full file-based recovery to be used a replica. This restriction should also apply to a copy with corrupted translog. If a copy with corrupted translog is mistakenly selected as a primary, it may not be able to provide all translog operations properly for replicas in the future.
This commit marks the store of a shard copy as corrupted if its translog is not good.

• on a contrast to Mark store as corrupted if its translog is bad #28434 translog corruption marker is stored in translog folder
• when we have translog corruption marker elasticsearch-shard remove-corrupted-data could be run without specifying any parameters (like index name and shard id) - it can scan all folders in data paths for the corruption markers

Relates to #31389

[elasticmachine]

Pinging @elastic/es-distributed