The use_state branch does not handle when __reduce__ returns a string

[liamhuber]

E.g. for the singleton pattern, as discussed in the pickle docs for __reduce__

MWE:

from abc import ABCMeta

import h5io

class Singleton(ABCMeta):
    _instances = {}

    def __call__(cls, *args, **kwargs):
        if cls not in cls._instances:
            cls._instances[cls] = super(Singleton, cls).__call__(*args, **kwargs)
        return cls._instances[cls]

class MySingleton(metaclass=Singleton):
    def __reduce__(self):
        return "my_global"  # The associated global variable

my_global = MySingleton()
my_global.data = "something"

h5io.write_hdf5(
    fname="stuff.hdf5",
    data=my_global,
    overwrite=True,
    use_json=False,
    use_state=True,
)

reloaded = h5io.read_hdf5(fname="stuff.hdf5")

assert reloaded is my_global
assert reloaded.data == "something"

This should re-import the singleton with its saved state, but instead raises an error ending in:

...
File ~/work/pyiron/h5io/h5io/_h5io.py:378, in _triage_write(key, value, root, comp_kw, where, cleanup_data, slash, title, use_json, use_state)
    366     reconstructor, _, state, *additional = reduced
    367 # For plain objects defining a simple __getstate__ python uses a
    368 # default reconstruction function defined in the copyreg module, if
    369 # an object wants to be reconstructed in any other way, we don't
   (...)
    374 # reconstruction of the object (documented in the pickle module)
    375 # that we don't implement currently in the _triage_read function
    376 is_custom = (
    377     reconstructor is not type(value)
--> 378     and reconstructor.__module__ != "copyreg"
    379 )
    380 if is_custom or len(additional) != 0:
    381     raise TypeError(
    382         f"Can't write {repr(value)} at location {key}:\n"
    383         f"Class {class_type} defines custom reconstructor."
    384     )

AttributeError: 'str' object has no attribute '__module__'