[BUG] Update with retry_on_conflict stays stuck in an infinite loop when sent as a part of bulk

[raghuvanshraj]

Describe the bug
When an update is sent as a part of bulk, the request stays stuck in an infinite loop in TransportShardBulkAction due to repeated retries. This code pointer is expected to limit the number of retries to retry_on_conflict specified by the user in the bulk request, but the retryCounter in the BulkPrimaryExecutionContext is never incremented in the resetForExecutionForRetry method. In a scenario where there are repeated conflicts for an update, the loop in TransportShardBulkAction remains stuck forever. Note that this same behaviour is not seen when the _update API is invoked and a single document is updated because for the _update API, retries are handled in TransportUpdateAction.

To Reproduce
We need to throw VersionConflictEngineExceptions repeatedly for this to show up. I have created a remote branch on my fork where I have modified the update code to always throw VersionConflictEngineExceptions: https://github.com/raghuvanshraj/OpenSearch/tree/retry-on-conflict-testing

Steps to reproduce the behavior:

1. Clone the branch linked above
2. Bring up the opensearch process
3. Create an index
4. Ingest a single document on the index
5. Update the document with the bulk API with retry_on_conflict set in the update request. Sample:

{ "update" : { "_index" : "{{index_name}}", "retry_on_conflict": 3, "_id": "IYGCtIsBspX0Krzt2kus" } }
{ "doc": { "counter": 3 } }

Expected behavior
The expected behavior in this case would be for retry_on_conflict to be honored and for the request to be succeeded/failed gracefully.

Plugins
NA

Screenshots
NA

Host/Environment (please complete the following information):
NA

Additional context
NA