an uncaught exception in captureError can result in infinite loop: can we guard that generally?

An uncaught exception in `captureError` can lead to an infinite loop between captureError ->
uncaught exception -> captureError, if `apm.captureError()` is called in a `process.on('uncaughtException', ...)`
handler. The agent does this itself if `captureExceptions: true`, which is the default.
https://github.com/elastic/apm-agent-nodejs/issues/2148 is an example of where this can happen.

This ticket is to think about if there is a general guard we could put on this. For example, could we use Async Local Storage to track if we are currently handling an uncaught exception, and if so, then abort starting another uncaught exception handler.