ENH: _lib: deobfuscate jax.jit crash in _asarray

[crusaderky]

>>> import jax
>>> import jax.numpy as xp
>>> from scipy._lib._array_api import _asarray
>>> a = xp.asarray([0])
>>> def f(a):
...     return _asarray(a, check_finite=True)
>>> jax.jit(f)(a)

Before:

ValueError: array must not contain infs or NaNs

After:

TracerBoolConversionError: Attempted boolean conversion of traced array with shape bool[].
The error occurred while tracing the function kmeans at /home/crusaderky/github/scipy/build-install/lib/python3.12/site-packages/scipy/cluster/vq.py:332 for jit. This concrete value was not available in Python because it depends on the value of the argument obs.
See https://jax.readthedocs.io/en/latest/errors.html#jax.errors.TracerBoolConversionError

Outside of JAX, it makes perfect sense to me that if you call _asarray(..., check_finite=True) on an array of strings you ought to get a meaningful TypeError.

[lucascolley]

makes sense, thanks!

[crusaderky]

Related: data-apis/array-api-compat#225