Fix Go http/grpc bad data

[grcevski]

I've made a few fixes and defensive changes with this PR:

• We weren't initializing some data structures not used by gRPC, which was printing garbage on the screen when print_traces is enabled.
• I moved reading of the method, path and content_length to the uprobe for ServeHTTP, just like we did it for the HTTP client. I believe this should resolve the issues https://github.com/grafana/beyla/issues/746 and https://github.com/grafana/beyla/issues/737.
• I added a defensive measure for the spans to check if they are valid and ignore them if not. This should prevent us from generating bad data if we don't update offsets on time.
• I resolved the issue for HTTP2 client.

Closes https://github.com/grafana/beyla/issues/737 Closes https://github.com/grafana/beyla/issues/746 Closes https://github.com/grafana/beyla/issues/742

These fixes need to be backported to 1.4.1.

The upgrade to http2 0.24 is purposefully not done to confirm our tests pass for the old version. We can then follow-up with a PR to upgrade our http2 dependency.

[mariomac]

Amazing!! I guess that so many years of coding in languages that automatically initialize the data to zero makes us to forget that C does not do it for us! 😅