Optimize data.heap

[shirok]

Looks like it has too much overhead.

http://chaton.practical-scheme.net/gauche/a/2017/04/11#entry-58ed2fb2-3245c https://gist.github.com/torus/35d71f5016957cd64659b0e5bb23ed21

[shirok]

Two significant factors:

1. Overhead of method dispatch: We use '~' a lot in data.heap so that the code works generically across different backend storage. It is order of magnitude slower than the specific referencer.
2. Overhead of comparator ops. Compared with direct comparison procedure, comparator ops (e.g. <?) are 7-10 times slower.

For the first point, we're introducing several measures to improve the performance. A couple of them are merged to master. More to come. https://github.com/shirok/Gauche/commit/55d275a78cee11d496d1b1f1ba62bccbd120b18d https://github.com/shirok/Gauche/commit/556fee985246eed0dd0189b3763d662bf0b0075b

For the second point, we've optimized comparator ops 2-3x so far. https://github.com/shirok/Gauche/commit/ac0441da21fc78024bb180394ebe3edb554b6d34