remove formatter function from percentage diff evaluation

[DJRickyB]

This commit fixes a small correctness issue with some metrics in the new Diff % column.

Before (sample):

|                                                        Metric |         Task |    Baseline |   Contender |         Diff |   Unit |     Diff % |
|--------------------------------------------------------------:|-------------:|------------:|------------:|-------------:|-------:|-----------:|
|                    Cumulative indexing time of primary shards |              |   0.0355667 |      0.0667 |      0.03113 |    min |      0.00% |
|             Max cumulative indexing time across primary shard |              |   0.0355667 |      0.0667 |      0.03113 |    min |      0.00% |
|                       Cumulative merge time of primary shards |              |     0.11245 |      0.0516 |     -0.06085 |    min |     -0.00% |
|                Max cumulative merge time across primary shard |              |     0.11245 |      0.0516 |     -0.06085 |    min |     -0.00% |
|                                                    error rate |       search |    0.655664 |           0 |     -0.65566 |      % | -10000.00% |

After (same sample):

|                                                        Metric |         Task |    Baseline |   Contender |         Diff |   Unit |   Diff % |
|--------------------------------------------------------------:|-------------:|------------:|------------:|-------------:|-------:|---------:|
|                    Cumulative indexing time of primary shards |              |   0.0355667 |      0.0667 |      0.03113 |    min |  +87.54% |
|             Max cumulative indexing time across primary shard |              |   0.0355667 |      0.0667 |      0.03113 |    min |  +87.54% |
|                       Cumulative merge time of primary shards |              |     0.11245 |      0.0516 |     -0.06085 |    min |  -54.11% |
|                Max cumulative merge time across primary shard |              |     0.11245 |      0.0516 |     -0.06085 |    min |  -54.11% |
|                                                    error rate |       search |    0.655664 |           0 |     -0.65566 |      % | -100.00% |

Relates #1333

[DJRickyB]

marked as :internal to keep it out of the release notes, minor bug fix for an as-yet-unreleased feature.

[DJRickyB]

saw this elsewhere, adopted here rather than making _diff() public

[pquentin]

Thanks! Now I'm curious however, why do we use convert.factor(100) for error rates?

[DJRickyB]

Now I'm curious however, why do we use convert.factor(100) for error rates?

We express our error rates as a percentage. It might be able to be refactored away from the formatter (so we just use the identity, the default) but I'm leaving that as out-of-scope for this change

[pquentin]

Thanks! Yes, definitely out of scope, I wasn't suggesting a refactoring.