This is a comprehensive refactoring of the error mechanism of the Intl extension.
By moving the prefixing of the current method/function being executed to actual error message creation by accessing the execution context, we get the following benefits:
- Accurate error messages indicating *what* call caused the error
- As we *always* "copy" the message, the `copyMsg` arg becomes unused, meaning we can reduce the size of the `intl_error` struct by 4 bytes.
- Saving it as a zend_string means we know the length of the message
- Remove the need to pass around a "function name" `char*` across multiple calls
- Use Intl's exception mechanism to generate exceptions for constructor call
- This removes the need for replacing the error handler
- Which didn't do anything anyway in silent mode, which required throwing non-descriptive exceptions
I'm not totally sure, but I have a strong suspicion that the fact
that this produces an error is an artifact of undefined cast behavior
(which will yield INDVAL on x86 but saturate on ARM). INF seems to
be the only value that results in an error even on x86 (variations
like -INF or NAN succeed).
It might make sense to just remove this test entirely, but for now
let's skip it on non-x86.
