h5repack tests for compression filters need review

[byrnHDF]

Adding the option of using zlib-ng exposed issues in the expectation of the h5repack tests. At least the compression tests should be reviewed as the data/files being used and the test options do not seem to be in sync.

In particular, the minimum size option should be skipping the application of filters when a dataset is smaller then the options value. Yet the file being used has datasets that are all larger and therefore the option does not get tested beyond the parsing of the command.