Add profiling to opentelemetry demo application (measure overhead)

While Pyroscope has benchmark suite available [here](https://github.com/pyroscope-io/agent-benchmarks) with [results here](http://pyroscope-agent-benchmarks.s3-website-us-east-1.amazonaws.com/latest.txt) , it is a somewhat naive approach to getting benchmarking data because those numbers are not really representative of a real application. If you profile a completely idle application it will look like profiling has a large "overhead" relative to profiling a more active/busy application.

It would be really nice to add Pyroscope profiling to the [open telemetry demo application](https://github.com/open-telemetry/opentelemetry-demo) for several of reasons:
- Grafana is already included in the demo-application and if configured correctly users could see profiles for each language in the dashboard
- We could also run eBPF profiling on the whole cluster and visualize it in Grafana
- We could use this to get benchmarks for relative overhead of profiling on a more realistic application
- We could use profiling to understand relative resource utilization of other signals such as logs / traces / metrics