Fix: respect config when opening an existing array

[ianhi]

Fixes #2931

Previously there were edge cases with open which would result in the write_empty_chunks config was being ignored. This PR has two separate fixes.
1st: for zarr.open line 327
2nd: zarr.open_array required some config parsing as well as a change to the api of AsyncArray.open

• Add unit tests and/or doctests in docstrings
• Add docstrings and API docs for any new/modified user-facing classes and functions
• [NA] New/modified features documented in docs/user-guide/*.rst
• Changes documented as a new file in changes/
• GitHub Actions have all passed
• Test coverage is 100% (Codecov passes)

[ianhi]

Discussed with @jhamman at the dev meeting and we came to the following conclusion:

1. Un-deprecate order and write_empty_chunks in favor of keeping these explicit
2. deprecate the config argument to create and to AsyncArray
3. default to the values in the global config when these values are None which should be the default

This keeps better consistency with zarr-python 2 and is more explicit in what the user is passing rather a layer of obfuscation by wrapping in ArrayConfigLike as a dict.

[ianhi]

Motivated by pydata/xarray#10177

cc @d-v-b

[d-v-b]

Why do our existing write_empty_chunks tests not work here?

[d-v-b]

for context, it would probably be preferable to expand the scope of an existing test instead of adding a new one

[ianhi]

Part of the fixes here ended up in #3378

but the remaining confusing with the possibliity of passing mulltiple ways of defining the same config remains