NH-12018 oboe Reporter init moved to Configurator

[tammy-baylis-swi]

This update moves oboe Reporter() from the SW Exporter to the SW Configurator so that oboe_init is called at custom-distro setup when using the SW Exporter or with a different exporter set by OTEL_TRACES_EXPORTER env var. There doesn't seem to be any other way to trigger oboe_init in the SWIG'd oboe_api, which makes sense as it looks like the only responsibilities of Reporter() are to assemble options then use them to oboe_init.

This also reverts a change made in #14 so that instead of load_entry_point for a non-SW exporter, we go back to:

exporter = next(
    iter_entry_points(
        "opentelemetry_traces_exporter",
        environ_exporter_name
    )
).load()()

I've done this because in my mind it's the cleanest way for custom-distro to find something like otlp_proto_grpc in the Python Distribution at runtime. It's too brittle to map required dist names for load_entry_point, and it's too complex and a waste to implement a search through sys.path when iter_entry_points already does it. Even though there is a small chance a customer might have multiples of the same environ_exporter_name in their path (see this comment), generally speaking that's for them to make sure is correct. Also, the iter_entry_points approach is used all over OTel Python and it's cleaner to match than to mix up setup techniques.

Special note: Currently OTel Python does not use entry points to configure Sampler like it does for Exporter and Propagators, so using iter_entry_points or load_entry_point for a non-SW Sampler fails. OTel Python uses sampling._get_from_env_or_default() so we do too: https://github.com/open-telemetry/opentelemetry-python/blob/main/opentelemetry-sdk/src/opentelemetry/sdk/trace/sampling.py#L374-L377 I've added this piece to the NH Python Troubleshooting doc.

Here is today's manual testing doc for this update: https://swicloud.atlassian.net/wiki/spaces/NIT/pages/3042050066/2022-05-04+configurator+testing

Please let me know what you think or if I've missed something 😃

[cheempz]

As always thank you for the informative PR writeup @tammy-baylis-swi! The changes LGTM with one caveat which is my fault for not having actually discussed it with you -- let's essentially pin our distro to ONLY use our custom sampler :). Added the rationale to your nice confluence page.

And indeed I see now there are several cases of the use of next(iter_entry_points().. outside of any type of looping in OTel Python, and it does make sense from the angle of working around the brittleness of specifying a dist :)

[tammy-baylis-swi]

let's essentially pin our distro to ONLY use our custom sampler :)

Thank you @cheempz ! Changed to always use our SW Sampler in 01f1e98. Setting OTEL_TRACES_SAMPLER now gets ignored.

I've also changed the Troubleshooting page back to "Not supported" for customizing OTel sampler. I kept my findings in a note in case we want to try something similar in the future.

Please let me know what you think 😺

[cheempz]

LGTM, thanks @tammy-baylis-swi! I've been glossing over how the agents should handle fatal misconfiguration but this PR has reminded me to jot it down in https://swicloud.atlassian.net/browse/NH-11751 so we can tackle that separately.