Fix masking for fill_value and missing_value for numpy 2.4.0

[valeriupredoi]

Description

I merged too fast #303 (like a .303 bullet hehe) but @davidhassell mentioned it's not complete since missing_value could be a multi-dim array, not just a size one, as I thought. This should fix that.

Reminder: numpy 2.4.0 doesn't allow stuff like ma.masked_equal or other such funcs that set fill_values, to have arrays passed for fill_value, eg their deprecation:

    @fill_value.setter
    def fill_value(self, value=None):
        target = _check_fill_value(value, self.dtype)
        if not target.ndim == 0:
            # 2019-11-12, 1.18.0
            warnings.warn(
                "Non-scalar arrays for the fill value are deprecated. Use "
                "arrays with scalar values instead. The filled function "
                "still supports any array as `fill_value`.",
                DeprecationWarning, stacklevel=2)
    
        _fill_value = self._fill_value
        if _fill_value is None:
            # Create the attribute if it was undefined
            self._fill_value = target
        else:
            # Don't overwrite the attribute, just fill it (for propagation)
>           _fill_value[()] = target
            ^^^^^^^^^^^^^^^
E           ValueError: setting an array element with a sequence.

This implementation here should be backwards compatible.

Before you get started

• ☝ Create an issue to discuss what you are going to do

Checklist

• This pull request has a descriptive title and labels
• This pull request has a minimal description (most was discussed in the issue, but a two-liner description is still desirable)
• Unit tests have been added (if codecov test fails)
• All tests pass

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 86.15%. Comparing base (451bab2) to head (571699a).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #304      +/-   ##
==========================================
+ Coverage   86.04%   86.15%   +0.10%     
==========================================
  Files           7        7              
  Lines         645      650       +5     
==========================================
+ Hits          555      560       +5     
  Misses         90       90              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[davidhassell]

Surely this only works when missing_value broadcasts to data, which in general it wont.

[valeriupredoi]

well I'm running out of any valid numpy function that can mask with an array then 😁

[valeriupredoi]

sorry, I got distracted by the news, how do you even apply a missing_value mask array if it's not broadcastable? I am really not familiar with anything missing value that's dim > 1 (or not scalar) - please suggest code, buds 🍺