Add support for FLAC compression of 64bit integers.

[tskisner]

• Add native support in the compiled extension for compressing 64bit integers with FLAC. The high and low 32bits are compressed in separate channels. On compression (for little-endian systems), the input 64bit integers are simply interpreted as interleaved stereo channels. On decompression, the separate channels in each frame are extracted to the high / low 32bit portions of the 64bit integers.

• Add 32bit / 64bit wrappers to the bindings, and support for 64bit integers to the mid-level encode / decode functions.

• Changes to the treatment of floating point data. float32 data is converted to int32 as before, but float64 data is now converted to int64 to gain additional precision.

• Changes to int64 encoding- previously this was restricted to either 32bit range around an offset or forced conversion as floating point data. Now this data is supported natively. Remove the int32/int64 conversion routines which are no longer needed.

• Change the HDF5 and Zarr on-disk format (bumping version to "1"). This change was necessary to support storing int64 data without an associated offset.

• Modify and expand unit tests to cover new int64 cases.

[tskisner]

This branch will be rebased after #8 is merged. I also want to expand the docs to describe the specific on-disk format. I also need to test the big-endian high / low 32bit handling in a multiarch qemu docker container.