B-trees supersede T-trees and are far more concurrent for multi-threaded, multi-core, multi-user data access and thus scale a lot better.
The above post is the only resource I could with much detail between the types. The TT documentation just says that B+tree indexes are the 'newer option that can offer concurrency benefits'. The two types of range index are not mentioned in the documentation's
In an OLTP environment you should definitely use B-tress (RangeIndexType=0). In a BI/Analytics environment (such as Exalytics) you should stick with T-Trees (RangeIndexType=1) for now as currently T-Trees are better optimised for BI workloads.