Sequence operation performance requirements for universal collection types

[johnwikman]

In PR #628, there was an identified need to rework the seqCmp to increase the performance for lists. However, as this change would make things slower for ropes, it was decided to not integrate this change into the standard library. Instead Daniel will use this optimization locally instead.

For the future, it would be good to have a system with the universal collection types such that the most "optimal" sequence operation (such as comparison) would be chosen depending on the underlying opaque data structure. I could see this being solved by either expanding the number of intrinsic (or built-in) operations for universal collection types or have a hint-based system. I suppose the latter would relate to previous discussions about making all intrinsic operations as external definitions, with sufficient higher-level hints to allow the compiler to perform the necessary optimizations.