feat: k6runner: improve error handling for k6 output

[roobre]

Closes #1060

This pull requests brings a number of improvements to how the local k6 runner, mostly used in private probe context, handles errors from k6. Most prominently:

• For errors that we can attribute to the script contents (user errors), it no longer returns nil, err (understood as "k6 could not be run"), but rather a response with a textual representation of the error, and the k6 output. This still causes the test to be marked as failed, but the error is reported to Loki and thus made visible.
• Logs are now parsed and searched for specific strings that, in SM context, are considered fatal errors.
• Limits are imposed on how many bytes will be read for log and metrics output. These are last-resort measures to prevent the whole payload to be rejected by mimir or loki if it is too big.

Based on #1061 as I want CI/CD to be able to run.