Very slow reads of datasets with compession by gathering or DSG ragged arrays

[davidhassell]

In versions 1.10.0.1 to 1.10.1.0, reading datasets with compression by gathering or DSG ragged arrays is very slow, taking seconds rather than milliseconds.

This is due to a bug in cfdm.Data._original_filenames, which causes every integer element in the list/count/index variable to be assessed for an original filename, which, of course, it does not have!

PR to follow.