BSLZ4 compression on pixel mask

In the Nov19 HDRMX meeting, there was significant interest in having the pixel mask and any other such datasets compressed, probably using BSLZ4.

Since we only use H5Dread to read the pixel mask, this adds a hard requirement on the user to have the correct filter on their $HDF5_PLUGIN_PATH. We might be doing people a favour by using the chunk-read strategy we use for the detector data, applying the decompression ourselves.