Publish benchmark jars

[rbasralian]

We should publish the jars from the benchmark repo as part of our release process, so that customers can use the same tooling to build their own benchmarks.

[stanbrub]

I think we should nail down at least what the ask is and what we are willing to do:

• Run our benchmarks with their hardware?
• Write/Run their own benchmarks?
• Run Groovy or Python benchmarks?
• Get only CSV output or publish to the cloud?
• Manage benchmark result history?
• Run against native DHC or Docker images?
• Notify results to slack or whatever?
• Run against DHE workers?
• Read from sources we don't?

[stanbrub]

Here's what we currently can do (support is another matter):

• Only runs against DHC, not DHE workers
• Only operational benchmarks are used in the standard runners
• Can run benchmarks from an IDE, from the command line, or from the GH workflows
• Only Python queries are supported in benchmarks
• Only running against DH docker images if DH restart is enabled
• Test runner is co-located with Deephaven service
• Running from GH workflows requires user-specific secrets for bare metal provisioning and GCloud publish
• Run from super-jar, which is bloated by Barrage client dependencies (122M)
• Ingest data from parquet, csv, kafka

[rbasralian]

The only explicit request from the customer is to publish jars for what currently exists (that is what we have advertised as existing, and since it is all that exists, it's all that they've seen). That said, they will certainly want to:

1. Run our existing benchmarks with their hardware
2. Write/run their own benchmarks (this is the main goal)
3. Use either Groovy or Python (preferably Groovy)
4. Get CSV output (publishing to cloud is not necessary; really just any output they can work from)

But the most important thing is making the tools we have now, in their current form, available to customers. They are likely to have additional requests down the road but it will be difficult to uncover those until they have something as a starting point.

We do not need to replicate our entire benchmark process/workflow; we just need to make it possible for them to create their own benchmarks and run them somehow (e.g. run a command with whatever manual or automated process they choose).