Improve monitoring errors

[lambdanis]

• Review tetragon_errors_total and other "errors" metrics
• Stop reporting non-errors (as in: no action needed) as errors. Define separate metrics for "casual fails" if needed.
• Refactor errors metrics to make them more useful (all of these are subject to review)
  • Delete tetragon_errors_total{type="handler_error"} - it duplicates handler_errors_total
  • Consider further splitting tetragon_errors_total
  • Consider merging event cache errors metrics into one with event_type and entry_type labels
  • Consider merging kprobe ok and errors metrics into one with status label
• Standardize on error label (rather than error_type, type, etc)