Feature/cc/do not write image

[wagoodman]

Pulling from @ChrisCinelli 's fork, this eliminates concurrency on loading, however, it fixes several bugs in the process (Fixes #105, #106 , #44 ). This also sets up for removing jotframe which will make dive cross platform and takes care of issue #95 .

[ChrisCinelli]

You anticipated me... since I wrote it already...
Before this PR dive was caching in a temp file the docker image. It was also allocating a few memory buffers that increased memory use and in case of large images created even more disk access because of memory being swapping out to disk by the kernel.

I have a lot faster performances on my machine after this change. Memory buffers have also be eliminated. This keeps the memory down even more and the CPU does not have to spend time copying memory around.

In order to achieve this I had to remove the parallel processing of layers. But, for what I can see, the code in this PR is a lot faster regardless and it does not wear out my SSD.

Because we do not use go routines anymore, the race condition on writing on layerMap is also removed that fixes #44.