Hickle

Minor release, two small bugfixes:
* Fix for 2D astropy `Time` objects from strings 162
* Fix for propagating `filename` arg 155

Minor release without dill dependency (for PyPi)

* Support for newer versions of numpy >= 1.21 and h5py >= 3.0
* Improved internal HDF5 structure for python dictionaries (no longer trailing `/data`)
* Deprecated use of `dill` in favor of in-built `pickle` (given the updates to pickle functionality in Py3, and issues with numpy dtypes)
* Switched to github actions for CI
* Objects referred to multiple times are now only dumped once within the HDF5 file ([HEP002](https://github.com/telegraphic/hickle/issues/139))

This is a minor patch release for v4 that fixes backward compatibility issues with v3 files.

This is a minor patch release for v4 that fixes a problem with hickling AstroPy `SkyCoord` objects.

*This is the major v4.0.0 release of the hickle package.*

**Changes:**
- Dropped support for Python 2.7;
- Dropped legacy support for `hickle` files made with v1 and v2;
- `OrderedDict` is now supported (65);
- Subclasses of supported classes can now be properly dumped;
- `data_0` is no longer used if there is only a single data group/set (44);
- HDF5 groups can now be dumped to and loaded from (54);
- Integers using Python's arbitrary-precision (integers larger than 64-bit) can now be dumped and loaded properly (113);
- Replaced broken link to `pickle` documentation with proper one (122);
- Objects that appear to be iterable are no longer considered as such unless `hickle` knows for sure they are iterables (70 and 125);
- Dict keys with slashes are now supported (124);
- Loaders are only loaded when they are required for dumping or loading specific objects (114);
- `hickle` now has 100% test coverage;
- NumPy arrays containing unicode strings can be properly dumped and loaded;
- NumPy arrays containing non-NumPy objects can be dealt with as well (90);
- Removed the use of 'track_times' (130);
- If an object fails to be hickled using normal means, `hickle` will now fall back to pickling the object;
- Massively simplified the way in which builtin Python scalars are stored, making it easier for the user to view.