feat(js/jvm): Emit stacktrace metrics

[loewenheim]

We never emitted stacktrace metrics (number of stacktraces/frames per request) for JS¹ and JVM. This adds the most important metrics for those platforms.

¹ Well we did for JS but under a different name.

[Swatinem]

code lgtm.

Though I would actually question whether these metrics are at all useful?
What conclusions do we draw from these? I believe the "number of" metrics are just a measure of how complex the events are. Do we even care about that?
The "unsymoblicated" / "missing sourcescontent" is also rather tied to customers misconfiguring their build system, or unsymbolicatable frames for which its not possible to access sourcemaps.

In either case, none of these metrics is anything we can active influence, or a measure of the quality of our service?

[loewenheim]

Though I would actually question whether these metrics are at all useful?

Right, fair question. I'm doing this in preparation for having Sentry send us platform values with events and stack frames. This is so that when we (eventually) get mixed stacktraces, we can differentiate the frames (recognizing native frames is easy, they're the only ones with an instruction_addr, but JS and JVM frames can't trivially be told apart).

I want to tag these metrics (specifically the num_frames) with the platforms they have set to see how often we actually have this information and whether e.g. JS events with non-JS platforms make it through.