Always run OTEL collector and optionally continue running after benchmarks

[lahsivjar]

PR implements the following improvements to the OTEL collector (with commit IDs for ease in review):

• Removes the unused -detailed flag (95a338f): -detailed flag was used in apm-server's bench test to print the detailed metric. In apm-perf/apmbench we print this by default.
• Always run the OTEL collector (5cf0a7d): before this PR, OTEL collector was run only when collector config were provided. These config specify which metric to store and without the config there was no point in running the collector. With this PR, we run the collector always in preparation for allowing another pipeline to ship all data to a persistent layer instead of only keeping the data in-memory.
• Allow apmbench to continue running otel collector (50c8a83): Introduces a flag -server-mode to indicate that collector can be kept running even after benchmark has finished. This also plays into adding a persistent pipeline in future.

[axw]

I don't understand the persistent pipeline bit, but it looks harmless and simple enough, so happy to merge and see where it goes :)