stdlibs benchmark regression

[Akirathan]

There are substantial regressions in some stdlib benchmarks that happened in some of these commits: 87291d5...9a9eff1

Regressions in:

• https://enso-org.github.io/engine-benchmark-results/stdlib-benchs.html#org_enso_benchmarks_generated_Table_Sorting_table_order_by_objects
  • 30x slowdown !!
  • 
• https://enso-org.github.io/engine-benchmark-results/stdlib-benchs.html#org_enso_benchmarks_generated_Map_Error_Benchmark_Slice_report_warning_Map_Id_All_Errors
• https://enso-org.github.io/engine-benchmark-results/stdlib-benchs.html#org_enso_benchmarks_generated_Map_Error_Benchmark_Slice_report_warning_Map_Id_No_Errors
• https://enso-org.github.io/engine-benchmark-results/stdlib-benchs.html#org_enso_benchmarks_generated_Map_Error_Benchmark_Slice_report_error_Map_Id_No_Errors
• https://enso-org.github.io/engine-benchmark-results/stdlib-benchs.html#org_enso_benchmarks_generated_Map_Error_Benchmark_Slice_ignore_Map_Id_All_Errors
• ...

[Akirathan]

Looking into the Table_Sorting.table_order_by_objects benchmark:

• It seems that the warmup is not sufficient, increased the warmup to 7 iterations per 7 seconds and measurement to 7 iterations per 7 seconds.
  • It seems that the benchmark is now unstable - in some executions I am seeing similar results to the old (good) revision, in some executions, I am seeing 2x slowdown.
  • Let's increase the warmup and measurement phases for now and analyze further in the future, if necessary.