Enhancement: add support for jemalloc

[dumblob]

On systems which support jemalloc natively (e.g. FreeBSD), we could try to use it due to its significant performance scalability gains in multithreaded applications. Even if we didn't want to support it in production for whatever reason, its profiling capabilities might still be worth supporting it for debugging and profiling purposes.

jemalloc can be used as a drop-in replacement, so we can just favour it when compiling.

We already discussed memory allocator(s) in https://github.com/daokoder/dao/issues/79, but this time, it's rather about support, not incorporation.