if only one fresh block left when kv_init, we should not just locate the workspace to this fresh one, otherwise we probably get a KV_ERR_NO_WRITEABLE_BLK in kv_init next time(if we write something to this workspace). We try a gc here to locate workspace to an inuse block rather than the fresh.
1. a true wear-leveling kv fs for norflash, especially optimize for some onchip norflash with "write once on one single write unit" like stm32l4, a true "no earse before write" flash algorithm.
2. an "as less as possible" gc strategy, do best to save norflash's life.
3. full "power down protection" support
4. see "examples" of kv, project in "TencentOS_tiny_EVB_MX_Plus", with onchip flash and qspiflash sample.
