Currently, variationIdMap is being populated as follows:

```
projectConfig.variationIdMap = {
      ...projectConfig.variationIdMap,
      ...keyBy(experiment.variations, 'id')
    };
```

This creates a new objects for each experiment unnecessarily. This causes large slowdown in project config parsing for datafiles containing a large number of experiments containing a large number of variations.

This commit makes this more efficient.