GitLab

There are changes with 1.12 which we can't use yet (need to maintain
compatibility down to 1.8 for now, but the subset of the HDF5 API we use
works with 1.8 and 1.10).

Otherwise, the cell comparison doesn't recognise the cells as the same,
breaking indexing with an R reference cell.

Unfortunately, the resulting cell after comparison by compare_r_c_p
still comes out as P, but that doesn't seem like a big loss.  To get it
strictly correct, we'd need some way of tracking through the information
that the cell got "uncentered" from R to P, even though the matrix is an
identity.

Makes it compatible with Python 3

Of course, these should never happen.

Enables compatability with zlib < 1.2.3.5

This was broken by d62367db, which was some enabling work for
reading geometry info from streams.

Replaced by cell_explorer many years ago

Normally the beam centre is masked out or not on the detector.  However,
when it is, pairing with 000 causes gradients to come out as NaN and the
prediction refinement to fail.

The units needed to be clarified.