Move benchmarking to OpenSearch Benchmarks

[jmazanec15]

Related issue: opensearch-project/opensearch-benchmark#103

OpenSearch benchmarks is the benchmarking tool used for the OpenSearch project. It includes a lot of functionality, including multiple clients as well as the ability to launch clusters.

Currently for this plugin, we use our own tool for benchmarking. This tool allows us to collect recall metrics as well as measure the query and index latency from data sets from bigannbenchmarks as well as annbenchmarks. This tool works well for those cases, but lacks a lot of the functionality that OpenSearch benchmarks has - for instance, load testing with multiple clients. Maintenance of the tool also requires a lot of duplicate effort.

OpenSearch benchmarks is very extendable. We should be able to create custom workloads that run the tests we are running with our own tool.

Once functionality is covered with OpenSearch benchmarks custom workloads, we can remove the benchmarking tool.