More efficient handling of long queries

[RagnarGrootKoerkamp]

As written in future work of the paper:

Following #4, for 100bp-1kbp queries, we may want to do some of the following:

• early abort on >k edits (Early break on distance >k #2). Maybe we can check after every 64 columns or so.
• Make the searched region more sparse, as in Fig 3b here.
• Do band-doubling or so for the traceback step.