It is sometimes useful to have the data in UTF-16, either because it cannot be represented in 8-bit or because a common use case is to return QStrings from them, which can then be fromRawData()'ed instead of requiring a heap allocation.

Acceptance criterion:

1. all existing tests still pass
2. the following new tests don't fail:

{
    constexpr auto table = qOffsetStringArray(u8"Foo", u8"Bar");
    QVERIFY(table.viewAt(0) == u8"Foo"); // returns QUtf8StringView
    QVERIFY(table.viewAt(1) == u8"Foo");
}
{
    constexpr auto table = qOffsetStringArray(u"Foo", u"Bar");
    QVERIFY(table.viewAt(0) == u"Foo"); // returns QStringView
    QVERIFY(table.viewAt(1) == u"Bar");
}

std::byte version's name to be bikeshedded.