std::bigint: improve performance

[gifnksm]

This commit reduces extra memory allocation in std::bigint.

In my simple benchmark test, time of computing 1000! becomes ~1000x faster.

• first benchmark result
• improved version's benchmark result

There still may be some points to be improved in performance. For example, switching the algorithms depending on the length of the array.

[catamorphism]

I would appreciate more comments in the vec_ops code that you're adding. Even though this is internal to the bigint library, somebody will have to maintain it.

Thanks!

[lomereiter]

Could you add figures for the GMP bindings available via cargo install gmp? Here's fibonacci and factorial using them: https://gist.github.com/4527144

[brson]

I agree this is a lot of tricky, uncommented code. Most people (like me) are not intimately familiar with bignum math and will not immediately understand what is going on here. @gifnksm can you document this code a little more? I don't even know what documentation it needs because I don't understand it.

[catamorphism]

I'm going to close this pull request and ask that @gifnksm submit a new one with comments if you're able to get around to it. Performance improvements are great, but they need to be documented.

[pnkfelix]

cc me. (i looked briefly at the bigint code we have now and may want to resurrect this patch, or at least review it before embarking on trying to improve our current bigint code.)