Add idempotency support for bulk request API

[dblock]

Is your feature request related to a problem? Please describe.

Coming from #3000, where we say current bulk indexing API places a high configuration burden on users today to avoid RejectedExecutionException due to TOO_MANY_REQUESTS. This forces the user to "experiment" with bulk block sizes, multi-threading, refresh intervals, etc.

Describe the solution you'd like

Make it easier for clients to handle retries by making bulk requests idempotent by generating a request ID on the client, or assuming id=SHA(data), tracking pending bulk write tasks on the server and their progress, making it possible to safely retry the entire bulk operation. Clients receiving a 429 would have a built-in retry, and it would also be safe to retry from multiple instances of the client with the hope that one succeeds eventually when the server finally has capacity.

Some googling shows people are at least attempting implementing client-side retry, https://gist.github.com/henrikno/e0ebd6804cb62491343c or https://gitlab.com/gitlab-org/gitlab/-/issues/12372

Describe alternatives you've considered
Streaming API, #3000.

[adnapibar]

There is a utility class for thread safe bulk processing with controls on when to flush, number of concurrent requests and retry policies - BulkProcessor.java

This can be used with a high level rest client, for example

try(RestHighLevelClient client = new RestHighLevelClient(RestClient.builder(new HttpHost("localhost", 9200)));
            BulkProcessor bulkProcessor = BulkProcessor.builder(
                (request, bulkListener) -> client.bulkAsync(request, RequestOptions.DEFAULT, bulkListener), listener)
                    .setBulkActions(500)
                    .setBulkSize(new ByteSizeValue(2L, ByteSizeUnit.MB))
                    .setFlushInterval(TimeValue.timeValueSeconds(10L))
                    .setConcurrentRequests(0)
                    .setBackoffPolicy(BackoffPolicy.constantBackoff(TimeValue.timeValueSeconds(1L), 3))
                    .build()) {
            bulkProcessor.add(new IndexRequest(...));
            ...

} catch (IOException e) {
          throw new RuntimeException(e);
}