Mark spoiled perf test results as spoiled or remove them.

[cijothomas]

https://open-telemetry.github.io/otel-arrow/benchmarks/continuous/

When we hit issues like #1549 that produces incorrect perf results, we need to find a way to

1. Either remove the results (with some maintainers-only workflow) so charts look clean
2. Annotate results with some text, so we can add an explanation/link-to-issue.