Unbalanced priority search pennants

The implementation in this section incorporates several small improvements over the one in Section 2.

• Lists have been replaced by John Hughes's efficient sequence type (in order to vaoid the use of `++').
• The keys and the priorities are stored directly in the nodes eliminating one level of indirection: we use Winnerkptm instead of Winner(k:->p)tm (see Remark 2 in the paper).
• The origin of a loser is encoded into the constructor of a loser tree: we use LLoser and RLoser instead of just Loser (see Remark 3 in the paper).
• The tournament view has been partly eliminated (in ordList, atMost, and atMostRange).