[BUG] Index state stuck in refreshing when Flint index deleted directly

[dai-chen]

What is the bug?

When users delete Flint index in OpenSearch directly by REST API, index refresh job exits but index state stuck in refreshing. In this case, recover statement won't take any action and users need to delete the state doc in metadata log or change state to empty manually.

How can one reproduce the bug?

Create a Flint index with auto refresh enabled. Delete the OpenSearch index directly and try to recreate it later.

What is the expected behavior?

Recover statement change the index state to EMPTY if index doesn't exist but index metadata log exists.