Make sampled transaction info to use flat buffers

Sampled transaction info is stored in native serialization format. But this data is read from other languages (Java, Python) and also need to support backwards compatibility. hence it is tedious to maintain this whenever there is a schema change. Since flat buffer support has been added, it would be a better idea to represent the transaction info data using flat buffers. This will ease the maintenance of this feature to support backwards compatibility and cross platform serialization/deserialization.

Start by locating the sampled transaction info serialization and the existing flat buffer support in the FoundationDB codebase. Verify how Java and Python consume this data and define compatibility coverage; done means cross-language serialization and deserialization remain backward compatible.

Written by the indexing model from the issue text.