Brute force & simple stupid concurrent MemTable

[rockeet]

Create an instance-thread-local Sub-MemTable for such MemTable, such MemTable is optimized for write and has read amplification.

• Each write thread write KeyValue into its Sub-MemTable
• Reader thread find or iterate for all Sub-MemTable

We first implement such MemTable on top of PatriciaTrie

• PatriciaTrie requires fixed memory size may be an issue
• realloc to shrink_to_fit, we assume realloc will not relocate memory when shrink.
  • jemalloc has function xallocx which is dedicated for resize memory in place

Once MemTable is not a bottle neck, the advantage of distributed compaction will be shown more...

[rockeet]

另一个问题：创建一个线程，写几条数据，然后退出线程，会导致数量巨大的 Sub-MemTable。 这个问题可以这样解决： 因为每个线程退出时，它对应的 instance_tls 对象会析构，在这个对象的析构过程中，我们可以做一些工作：如果是第一次析构（第一个线程退出），啥都不干，如果是后续其它线程的析构，就尝试与之前析构的线程对应的 instance_tls Sub-MemTable 进行合并。

------ 这个问题优先级较低，先实现其他功能

[rockeet]

Done！