add incremental search at least as an option

[polm]

add incremental search at least as an option

[polm]

Got this working but ran into some issues.

A big one is that while edict2 has many more entries than the SKK dict used as a base by other migemos, a lot of those entries are pretty redundant - 東京発, 東京行き, 東京着 for example. This means that the pruning step has more work to do, making things slower.

I think I'm going to put this on the backburner until someone requests it and try to find other approaches; even in the good existing migemos things are weird when your input is something short like "to" for example.

[polm]

Fiddling around with building a prefix tree things were much faster than I expected but still slower than is good. However, by caching single kana (as well as clusters like きょ) conversion became much faster with a minimal increase in dictionary size (5.1MB -> 5.4MB uncompressed). There is still noticeable lag on a few combinations and particularly on deletion, so caching should be added.

[polm]

Caching did the job; interactive speed feels pretty acceptable now. Load time is a bit longer than I'd like but not a big deal.