MNT Fix unit tests

[GuySartorelli]

Fixes various date formatting failures in https://github.com/silverstripe/silverstripe-framework/actions/runs/11091228469/job/30830628977

Two main causes:

1. IntlDateFormatter output changes based on operating system used php/php-src#16127
   • Bug in upstream dependency results in some "medium" dates being returned as "short" dates for en_NZ locale. We shouldn't try to pre-fix the bug, nor should we attempt to revert to the buggy behaviour. Solution is to test against a locale that didn't exhibit the bug.
2. strtotime returns false when string includes narrow non breaking space character php/php-src#11600
   • Intentional change in upstream dependency results in "narrow non-breaking spaces" in some formatted dates where a regular space used to be. Solution is to update tests to not care what the space looks like so long as it's there.

Issue

• Broken builds and merge-ups .github#313

[GuySartorelli]

The old date is effective the same as you'd get when using IntlDateFormatter::SHORT, though we're using IntlDateFormatter::MEDIUM here.

Steve and I had talked in person about making the default configurable.
That would be doable for DateField_Disabled in the above test, because the default for that is returned from DateField::getDateLength() as a fallback value. Though its notable that none of the tests for DateField itself failed... so I'm not sure about the value of making that change but can do if the reviewer wants me to.

That's not possible for DBDate in this test though, because the default value is a default argument in the method signature of DBDate::getFormatter() and we can't change method signatures in patches or minors.

[emteknetnz]

Though its notable that none of the tests for DateField itself failed

Looks like that's because only the internal values were being validated in DateFieldTest, rather than the formatted values like Nice()

so I'm not sure about the value of making that change but can do if the reviewer wants me to.

Reading php/php-src#16127 (comment) particularly around the "no contract" bit, I guess it's correct to not have config here and not attempt to retain backwards compatibility. Simply updating our unit tests is the correct to test the new format is the correct response I think.