Added a retry logic and a service availability check function for high availability.

[OlehPalanskyi]

Description

Problem:
If there are two servers in the hosts field (example: hosts "server1:9200,server2:9200"), automatic switching to the available host does not work. If server1 is unavailable, it does not switch to server2, and the service writes an error and freezes:
"Unexpected error raised. Stopping the timer. title=:in_opensearch_timer error_class=Faraday::ConnectionFailed error="EOFError (EOFError)""

Solution:
I added a retry logic in case of connection failures.
I created a service availability check function for high availability. New function get_reachable_hosts in in_opensearch.rb
I used one library opensearch to solve this problem.
And added ignoring any clear_scroll errors.

Checklist

• tests added
• tests passing
• README.OpenSearchInput.md updated
• History.md and version in gemspec are untouched
• backward compatible

[cosmo0920]

I left some of the my concerns. No worries, the direction of the patch should be reasonable but we need to handle for error cases and retrying period precisely.

[OlehPalanskyi]

Thank you for your comments
I'll rework my patch based on your suggestions.

[cosmo0920]

Thanks for your work. It's brilliant work! All of the implementation should be exceeded my expectations.

[cosmo0920]

@kenhys Any concerns?

[cosmo0920]

On thing, I had a nitpick though. The retry_state often should be used inside of <buffer> sections. However, Input plugins of Fluentd do not have buffering mechanism. So, it should be wrong usage if we use <buffer> for this retrying functionality.
How you do think this nitpick, @kenhys or @daipom ?
The main concern is including retrying configurations should be going to some of the messing up for configuration parameters. So, we need to handle them within the dedicated sections somehow but it's not mandatory but optional thing.