Add benchmarks for the Observation API

[jonatan-ivanov]

Add benchmarks so that we can compare the Observation API with other ways of recordings.

• Add benchmarks for comparing the Observation API with Timer + LongTaskTimer
• Add benchmarks for NoopObservation and an uninstrumented case
• Add benchmark for ObservationOrTimerCompatibleInstrumentation
• Use @BenchmarkMode(Mode.SampleTime)
• Add different scenarios to test this with and without contention (one/multiple threads)
• Include the GC profiler so that we can see how much garbage is generated
• Check if we can run these against some existing instrumentation where we can use fakes (i.e.: no outgoing call); this might be more realistic