Test capture url at startup and print error message if not valid

[robfromboulder]

For testing and general use, it would be super helpful if the logger validated the capture url by sending a ping to the remote service.

The logger fails if the capture url is not set at all, but there's no warning today if the capture url doesn't resolve on the network. This could be caused by a DNS issue from the target host, or human error (like copying the capture url from the documentation without replacing the example host name), or just a typo.

It would save time and reduce documentation if the capture url were actively validated at startup. In the case where API calls are not being logged, we currently describe this as either a problem with the eBPF probes or a networking problem (and that the networking config should be double-checked manually). By actively pinging the capture url, we could simplify troubleshooting and eliminate the manual check.

We will have to decide what we want the ping request to look like, so that this returns an expected value to the logger, without actually writing the ping request/response to the database.

[robfromboulder]

Logger change is done but need to update README and TESTING accordingly

[monrax]

Just updated TESTING. Please, take a look when it is convenient.

I could only find one mention of this networking config in the README. Should we remove that line from the System Requirements?

logger-ebpf/README.md

Line 33 in ddc4357

* Network access to the Kubernetes cluster where [Graylog API Security](https://graylog.org/products/api-security/) is running

In addition to the fluke reachability check at startup, I added another check before the logger submits a new message bundle. Not sure if effectively duplicating the number of requests (even though the payload is empty when pinging) could be overkill. We could also introduce a check frequency, to perform the reachability check every X number of messages, or just get rid of the additional checks, if so.