Enable `QuadPrecision()` operating on `np.ndarray` or `np.number`

[dlakaplan]

Right now to convert a scalar to a quad precision you can do:
QuadPrecision(x)

but for a np.ndarray or np.number you need to do:
x.astype(QuadPrecDType)

It would be nice if they had the same interface.

[ngoldbaum]

This is how numpy's builtin scalars work:

>>> import numpy as np
>>> np.float64([3, 4])
array([3., 4.])
>>> np.float64(np.array([3, 4]))
array([3., 4.])

But for QuadPrecision:

>>> QuadPrecision([3, 4])
Traceback (most recent call last):
  File "<python-input-5>", line 1, in <module>
    QuadPrecision([3, 4])
    ~~~~~~~~~~~~~^^^^^^^^
TypeError: QuadPrecision value must be a quad, float, int or string, but got <class 'list'> instead
>>> QuadPrecision(np.array([3, 4]))
Traceback (most recent call last):
  File "<python-input-6>", line 1, in <module>
    QuadPrecision(np.array([3, 4]))
    ~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^
TypeError: QuadPrecision value must be a quad, float, int or string, but got <class 'numpy.ndarray'> instead

Maybe we can call np.array if we're handed a sequence. Ideally we can support this without reimplementing all of numpy's recursive array coercion logic.

[dlakaplan]

Yes, that would also help with lists

[SwayamInSync]

Thanks, that should be straightforward to implement and I think PyArray_FromAny can avoid doing all the recursive coercion.