Define observability interface between core SDK and the wrapper

[avsej]

No description provided.

[avsej]

@DemetrisChr, @thejcfactor
I think it is mostly ready for review and testing.

1. Bridge interface sample (for wrappers): --write-telemetry-to-file option of cbc and core/file_signal_sink.cxx/core/file_signal_sink.hxx

./build/tools/cbc get --use-replica=all --write-telemetry-to-file=./telemetry.json foo

2. OpenTelemetry integration (for public API): --tracing-use-opentelemetry/--tracing-opentelemetry-* and --metrics-use-opentelemetry/--metrics-opentelemetry-*:

./build/tools/cbc get --use-replica=all --tracing-use-opentelemetry --tracing-opentelemetry-sampler=always_on foo
./build/tools/cbc  pillowfight --metrics-use-opentelemetry

[DemetrisChr]

What would the consumers' response be when they receive an std::nullopt result from here? My understanding is that they would call take_buffer in that case. Wouldn't it make sense to return the buffer if that's what they would do? Or do we want to give the consumers more flexibility?

[avsej]

The logic might be different on whether the buffer is empty and when condition variable waiting timeout happened. For example, the wrapper might have its own sleep or yield the processor if it is empty but not timeout.