Prevent beyla self-instrumentation

[grcevski]

Depending on how customers choose to deploy beyla, we can easily end up in a situation where Beyla is self-instrumenting itself. One example would be setting "all" for an executable name or a kubernetes pod/service/namespace. This is particularly common actually, because our default helm configuration instruments all namespaces, so Beyla will self-instrument even when it's installed in its own namespace.

Self-instrumentation comes with serious consequences, mainly because of self-perpetuating instrumentation of calls to /v1/metrics and /v1/traces. This creates unnecessary load on all components in the system and drowns the collection pipeline.

I've added an option to purposefully enable Beyla self-instrumentation, however I think we shouldn't even document it. It should be purely internal option we may choose to enable in some special circumstances.

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 81.50%. Comparing base (f9a8d4b) to head (f245fec).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1048      +/-   ##
==========================================
- Coverage   81.50%   81.50%   -0.01%     
==========================================
  Files         139      139              
  Lines       11113    11119       +6     
==========================================
+ Hits         9058     9062       +4     
  Misses       1543     1543              
- Partials      512      514       +2     

Flag	Coverage Δ
integration-test	56.91% <100.00%> (+0.04%)	⬆️
k8s-integration-test	58.93% <50.00%> (-0.11%)	⬇️
oats-test	37.42% <50.00%> (+<0.01%)	⬆️
unittests	50.99% <33.33%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[mariomac]

Good feature!