Application-level profiling

There should be some application-level API (i.e., a Manager) available for reporting data that will be included with the existing profiling stream. This is especially useful for doing things like correlating application events (e.g., a given kind of monster was killed) with system details (e.g., transaction timeout).

This can be prototyped without any changes to the core, but in order to report the kinds of data types that application developers will likely find useful (strings, enums, etc.) we will need to add some fields to the profiling reports. We should also keep in mind the kind of debugging tools that we will want to use this, and how those interfaces may affect this feature.