Release b2h5py v0.3.0.

This release changes the package interface to require explicitly enabling
optimizations and use more friendly names.

Functions for doing, undoing and checking patching are now called
`b2h5py.enable_fast_slicing()`, `b2h5py.disable_fast_slicing()` and
`b2h5py.is_fast_slicing_enabled()`, while the context manager is called
`b2h5py.fast_slicing()`.  Please see the updated and fixed example script.

Importing `b2h5py` no longer enables optimizations until
`b2h5py.enable_fast_slicing()` or `b2h5py.fast_slicing()` are used.  For the
previous implicit behaviour you may just import the new `b2h5py.auto` module.

Finally, the caching of the property that determines whether a dataset is
amenable for Blosc2 optimized slicing has been fixed.

pyproject.toml

@@ -20,7 +20,7 @@ classifiers = [
 authors = [
     { name = "Ivan Vilata-i-Balaguer", email = "[email protected]" },
 ]
-version = "0.2.1.dev0"
+version = "0.3.0"
 requires-python = ">= 3.3"  # contextlib.ExitStack
 dependencies = [
     "blosc2 >= 2.3.0",  # open schunk by offset