Add support for ignoring duplicate write tuple keys and missing delete tuple keys

[B3zaleel]

Checklist

• I agree to the terms within the OpenFGA Code of Conduct.

Describe the problem you'd like to have solved

According to the OpenFGA API docs, and after testing this locally, the OpenFGA API supports setting "on_duplicate": "ignore" within the writes object and "on_missing": "ignore" on the deletes object.

Could we add support for this as well?

Describe the ideal solution

Client Write and Delete options could support these values, which would then be added in the writes and deletes object.

Users spend a lot of time working around errors resulting from writing an already existing tuple or deleting a non-existent one.

Describe the ideal solution

OpenFGA v1.10.0 introduced support for write conflict options (on missing/on duplicate).

The SDKs must expose that in the underlying API client and models, as well as add support for idempotency options in the SDK client interface.

As an example for Go (each SDK should follow it's own similar conventions), the client should support the following options through ClientWriteOptions.Conflict:

• OnDuplicateWrites: Controls behavior when writing a tuple that already exists
  CLIENT_WRITE_REQUEST_ON_DUPLICATE_WRITES_IGNORE - Silently skip duplicate writes
• CLIENT_WRITE_REQUEST_ON_DUPLICATE_WRITES_ERROR - Return error on duplicates (default)
  OnMissingDeletes: Controls behavior when deleting a tuple that doesn't exist
• CLIENT_WRITE_REQUEST_ON_MISSING_DELETES_IGNORE - Silently skip missing deletes
• CLIENT_WRITE_REQUEST_ON_MISSING_DELETES_ERROR - Return error on missing deletes (default)

options := client.ClientWriteOptions{
    Conflict: client.ConflictOptions{
        OnDuplicateWrites: client.CLIENT_WRITE_REQUEST_ON_DUPLICATE_WRITES_ERROR,
        OnMissingDeletes: client.CLIENT_WRITE_REQUEST_ON_MISSING_DELETES_IGNORE,
    },
}

If the user does not set the default, the SDK should not send it, relying on the server to implement its default behaviour.

Ensure that the:

• Change has tests for both the OpenFgaApi and OpenFgaClient layers
• The Write section in the README is appropriately updated
• The unreleased section of the Changelog is updated

Alternatives and current workarounds

No response

References

No response

Additional context

No response