Fixes #49

I also included some small tests.

One could most definitely choose to use the same method as in RmwNorFlashStorage::write of splitting the range into many small blocks, but I've found in my project that small reads are quite detrimental when the flash instructions have to transit over some channel, like SPI.

In my project, I went with a similar solution of doing two small reads at the head and at the tail of the requested range, and then one big read for the main body, amortizing the runtime cost of individual reads.

I wasn't sure if the read in RmwNorFlashStorage::write was unaligned, but if it happens to be unaligned, then the subsequent calls to erase/write would fail anyway.