test_macroedge(tet) fails on some hppa machines

basix is failing `test_macroedge[1-CellType.tetrahedron]` on some debian hppa machines.
Build logs can be found at https://buildd.debian.org/status/logs.php?pkg=fenics-basix&arch=hppa

When failure occurs, the message is
```
____________________ test_macroedge[1-CellType.tetrahedron] ____________________

cell_type = CellType.tetrahedron, order = 1

@pytest.mark.parametrize(
"cell_type",
[
basix.CellType.interval,
basix.CellType.triangle,
basix.CellType.tetrahedron,
basix.CellType.quadrilateral,
basix.CellType.hexahedron,
],
)
@pytest.mark.parametrize("order", [0, 1, 2, 3, 4])
def test_macroedge(cell_type, order):
if cell_type == basix.CellType.triangle and order > 2:
pytest.xfail("Degree > 2 edge macro polysets not implemented on triangles.")
if cell_type == basix.CellType.tetrahedron and order > 1:
pytest.xfail("Degree > 1 edge macro polysets not implemented on tetrahedra.")

Qpts, Qwts = basix.make_quadrature(
cell_type, 2 * order + 1, polyset_type=basix.PolysetType.macroedge
)
basis = basix.polynomials.tabulate_polynomial_set(
cell_type, basix.PolysetType.macroedge, order, 0, Qpts
)[0]

ndofs = basis.shape[0]
mat = np.zeros((ndofs, ndofs))
for i in range(ndofs):
for j in range(ndofs):
mat[i, j] = sum(basis[i, :] * basis[j, :] * Qwts)
> assert np.allclose(mat, np.eye(ndofs))
E assert False
E + where False = (array([[ 1. , 0. , 0. , 0. , 0. ,\n 0. , -0. , 0. ... , -0. , -0. , -0.1711633 ,\n -0.2292677 , 0.26957468, 0.0286095 , -0.12574794, 0.890625 ]]), array([[1., 0., 0., 0., 0., 0., 0., 0., 0., 0.],\n [0., 1., 0., 0., 0., 0., 0., 0., 0., 0.],\n [0., 0., 1., ...0., 0., 1., 0., 0.],\n [0., 0., 0., 0., 0., 0., 0., 0., 1., 0.],\n [0., 0., 0., 0., 0., 0., 0., 0., 0., 1.]]))
E + where = np.allclose
E + and array([[1., 0., 0., 0., 0., 0., 0., 0., 0., 0.],\n [0., 1., 0., 0., 0., 0., 0., 0., 0., 0.],\n [0., 0., 1., ...0., 0., 1., 0., 0.],\n [0., 0., 0., 0., 0., 0., 0., 0., 1., 0.],\n [0., 0., 0., 0., 0., 0., 0., 0., 0., 1.]]) = (10)
E + where = np.eye

test_orthonormal_basis.py:164: AssertionError
=========================== short test summary info ============================
FAILED test_orthonormal_basis.py::test_macroedge[1-CellType.tetrahedron] - as...
= 1 failed, 9998 passed, 337 skipped, 72 deselected, 23 xfailed in 40618.78s (11:16:58) =
```

The failure is occuring on Debian's p-machines (pasta, pacific, pad), but passing on s (sap). That suggests to me it's a fault with some subtypes of hppa.

Perhaps it's a bug in gcc, rather than some faulty type assumption in basix.