[BUG] MeanViews has unexpected behavior for 0-D entries

[yimuchen]

Describe the bug

When reducing a 1-D histogram to 0-D via a hist.Hist.profile call, the resulting MeanView does not handle method calls as expected:

Steps to reproduce

import hist 
import numpy as np
h = hist.Hist(hist.axis.Regular(10, -10, 10, name='x'))
h.fill(np.random.normal(size=1000))
profile = h.profile('x').view()
print(profile)           # MeanView: (count, value, _sum_of_deltas_squared)
                         # (1000., 0.02, 1302.2964)
print(profile.item()[1]) # 0.02  
profile.mean()
profile.variance()

The last 2 line raises the following error message:

• For profile.mean():

TypeError                                 Traceback (most recent call last)
File /opt/conda/lib/python3.8/site-packages/numpy/core/_methods.py:179, in _mean(a, axis, dtype, out, keepdims, where)
    [176](file:///opt/conda/lib/python3.8/site-packages/numpy/core/_methods.py?line=175)         dtype = mu.dtype('f4')
    [177](file:///opt/conda/lib/python3.8/site-packages/numpy/core/_methods.py?line=176)         is_float16_result = True
--> [179](file:///opt/conda/lib/python3.8/site-packages/numpy/core/_methods.py?line=178) ret = umr_sum(arr, axis, dtype, out, keepdims, where=where)
    [180](file:///opt/conda/lib/python3.8/site-packages/numpy/core/_methods.py?line=179) if isinstance(ret, mu.ndarray):
    [181](file:///opt/conda/lib/python3.8/site-packages/numpy/core/_methods.py?line=180)     ret = um.true_divide(
    [182](file:///opt/conda/lib/python3.8/site-packages/numpy/core/_methods.py?line=181)             ret, rcount, out=ret, casting='unsafe', subok=False)

TypeError: cannot perform reduce with flexible type

• For profile.variance():

File /opt/conda/lib/python3.8/site-packages/boost_histogram/_internal/view.py:243, in MeanView.variance(self)
    [240](file:///opt/conda/lib/python3.8/site-packages/boost_histogram/_internal/view.py?line=239) @property
    [241](file:///opt/conda/lib/python3.8/site-packages/boost_histogram/_internal/view.py?line=240) def variance(self) -> "np.typing.NDArray[Any]":
    [242](file:///opt/conda/lib/python3.8/site-packages/boost_histogram/_internal/view.py?line=241)     with np.errstate(divide="ignore", invalid="ignore"):
--> [243](file:///opt/conda/lib/python3.8/site-packages/boost_histogram/_internal/view.py?line=242)         return self["_sum_of_deltas_squared"] / (self["count"] - 1)

File /opt/conda/lib/python3.8/site-packages/boost_histogram/_internal/view.py:18, in View.__getitem__(self, ind)
     [16](file:///opt/conda/lib/python3.8/site-packages/boost_histogram/_internal/view.py?line=15) # If the shape is empty, return the parent type
     [17](file:///opt/conda/lib/python3.8/site-packages/boost_histogram/_internal/view.py?line=16) if not sliced.shape:
---> [18](file:///opt/conda/lib/python3.8/site-packages/boost_histogram/_internal/view.py?line=17)     return self._PARENT._make(*sliced)  # type: ignore[attr-defined, no-any-return]
     [20](file:///opt/conda/lib/python3.8/site-packages/boost_histogram/_internal/view.py?line=19) # If the dtype has changed, return a normal array (no longer a record)
     [21](file:///opt/conda/lib/python3.8/site-packages/boost_histogram/_internal/view.py?line=20) if sliced.dtype != self.dtype:

TypeError: iteration over a 0-d array

It looks like special case handling needs to be done for 0-D results for View.