Fixed the silent-batch-drop bug where OpenSearch bulk indexing fails during shard reallocation

[todvora]

Description

During a rolling restart, OpenSearch can return a bulk response whose _shards.failures[] entry omits the shard number while a shard is still being promoted; the opensearch-java 3.x client treats that field as required and throws MissingRequiredPropertyException (see opensearch-java#551, fix proposed in #2023 but not yet merged), which didn't match any condition in Messages#createBulkRequestRetryerBuilder()'s retry predicate and caused the entire in-flight batch to be dropped instead of retried. MessagesAdapterOS now detects this specific failure (missing shard on ShardSearchFailure, to avoid masking unrelated parsing bugs) and converts it into a new IncompleteBulkResponseException, which Messages retries with the same exponential backoff as other transient indexing failures until the shard reallocation completes.

Motivation and Context

Fixes #26853

How Has This Been Tested?

Added unit tests

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Refactoring (non-breaking change)
• Breaking change (fix or feature that would cause existing functionality to change)

Checklist:

• My code follows the code style of this project.
• My change requires a change to the documentation.
• I have requested a documentation update.
• I have read the CONTRIBUTING document.
• I have added tests to cover my changes.

[moesterheld]

lgtm. for a backport, we will take into account the retyping from ShardFailure there to ShardSearchFailure