Add ZSTD Compressor

[TodicaIonut]

This PR adds a new compression algorithm optimized for Deep images.
During the development, deep images were shrunk by 20-30% compared to ZIPS with similar compression performance and 35-45% smaller when the compression level is turned to the max (about a 3-4x slowdown compared to ZIPS). The decompression speed slightly faster than ZIPS.

I did not spend a lot of time trying to optimize for non-Deep images, but
certain small non-deep example images with HALF pixels yielded a slightly worse compression rate compared to ZIPS.

The core algorithm is implemented in Blosc, with ZSTD codec with an appropriate data prefiltering.
The main advantage is that it offers a unified API access to multiple codecs and is under active development. Also the binary serialization stays compatible if ever we want to change the codecs/config parameters in the future, without introducing breaking changes.

Open questions:
I still have not figured out how to integrate Blosc properly into the OpenEXR build pipeline.

zstd support has not yet been merged into OpenEXR. cc @pleprince @clusty

I created the branch and called it zstd-beta.

[lgritz]

You know, prior to submitting a PR, you can just push to your fork and have it run CI and iterate until you have everything working, and then submit the PR.