schema evolution of event object types

[tomeichlersmith]

We achieve schema evolution by introducing two new features.

1. Persisting of data "structure" where type and version attributes are given to every H5 object in the event object recursion tree so that the versioned type by which an object was serialized can later be read.
2. Allow the objects being attached by a user class to either be only-loaded, only-saved, or both. This is the minimum structure necessary to rename variables on disk in order to update the schema or to calculate a variable in a different way.

To Do

• [x] Update I/O documentation to explain the details of this big update as well as include examples for users want to evolve their own data classes
• [x] Update Event and Run Headers to use this method when reading from ROOT?
• [x] Basic testing of schema evolution is implemented, but need to think about if there is more to do

[tomeichlersmith]

First round of benchmarking results shows no change when updating to handle schema evolution. It is important to note that the benchmarking does not test the speed of reading an old version with a new version; however, it does test the time necessary to serialize the extra type/version attributes.