Handling very large traces

[jfdube]

When dumping very large traces, it would be useful to implement a simple string dictionnary at the beginning of the json file, and every named event would reference a string in the dictionnary, maybe by index. This would greatly reduce the size of the json file.