Calling `exporter.flush` after fork to clear any existing profiling
data works, but it's rather heavy-handed and can generate a few
confusing log messages:

```
DEBUG (dd-trace-rb/lib/datadog/profiling/scheduler.rb:73:in `after_fork') Flushing exporter in child process #after_fork and discarding data
DEBUG (dd-trace-rb/lib/datadog/profiling/stack_recorder.rb:25:in `serialize') Encoded profile covering 2022-11-10T09:42:22Z to 2022-11-10T09:42:22Z
DEBUG (dd-trace-rb/lib/datadog/profiling/exporter.rb:52:in `flush') Skipped exporting profiling events as profile duration is below minimum
```

To improve this, I've cleared a new `Exporter#clear` that is still
(currently) implemented using `flush`, but now the exporter knows
what's going on more explicitly, and so it can implement `clear`
in a better way.

…eeded

The previous `flush` call was a placeholder, and we can now improve it.

This will be used by the `Exporter` when needing to clear data without
serializing it.

As the `StackRecorder` implements `#clear` natively, no need to
simulate it with a call to `#serialize`.

I decided not to bother updating the `OldRecorder`, so for that
class we'll still use the old behavior.