Align arg types of H5D_chunk_iter_op_t with H5Dget_chunk_info

[mkitti]

Fix #2056

For H5D_chunk_iter_op_t

• filter_mask changes from type uint32_t to unsigned
• size changes from uint32_t to hsize_t

The practical change is that size argument is now a 64-bit unsigned integer.

These changes make H5D_chunk_iter_op_t similar the argument similar to the size argument in H5Dget_chunk_info.

H5_DLL herr_t H5Dget_chunk_info(hid_t dset_id, hid_t fspace_id, hsize_t chk_idx, hsize_t *offset,
                                unsigned *filter_mask, haddr_t *addr, hsize_t *size);

The change is important because it allows the internal chunk size to change to be a 64-bit value in the future without modifying the public API.

[derobins]

The change is important because it allows the internal chunk size to change to be a 64-bit value in the future without modifying the public API.

The problem with this, though, is that using a 64-bit value now makes the function signature misleading until we update the file format. That will definitely not happen in 1.14.0, so we should save this change until HDF5 2.0 (or whenever we get around to updating the file format).

[mkitti]

I find the inconsistency with the rest of the public chunk API more problematic at the present.

The mismatch between hsize_t and the current maximum chunk size has existed for as long as hsize_t became a 64-bit integer.

[derobins]

I'll approve this and we can merge this in, but:

(A) This HAS to go to 1.12 ASAP since the chunk iteration stuff was merged there (but not released in 1.12.2).
(B) I'm not excited by the types in these functions and we should update them in 2.0 with a chunk dimension type and a fixed-size mask type.

[mkitti]

I rebased on develop since it's been a while.

[mkitti]

I cherry picked this on the hdf5_1_12 branch: #2329