Check performance of the tensor keys

[terrylyons]

These are used very often; particularly in sparse operations and the shuffle algebras. The performance has been improved in several of the tensor key operations, particularly concatenation, by better code (Thanks SAM!). It would be worth running a profiler and checking which operations consume the time, as some others could be better optimized if this was useful to the overall process. _tensor_basis.h