Fix String::begins_with when both strings are empty

[miv391]

There are two versions of String::begins_with:

bool String::begins_with(const String &p_string) const
bool String::begins_with(const char *p_string) const

In master, they work differently when checking if "".begins_with("").

String empty("");
bool b1 = empty.begins_with(empty);
bool b2 = empty.begins_with("");

b1 will be true, but b2 will be incorrectly false.

This PR adds more tests for begins_with and fixes the version using a const char * parameter.

I also added tests for cases where the parameter is null. Null parameter is possible only with the version using const char * parameter. anything.begins_with(nullptr) is expected to return false.

I added more tests also for ends_with and made some simplifications for the test case.

It is possible that this breaks something if there exists code that expects "".begins_with("") returning false.

[Mickeon]

Would be interesting to see where the const char * is used if easily possible just to ensure nothing breaks (which I'm doubtful of).

[miv391]

Would be interesting to see where the const char * is used if easily possible just to ensure nothing breaks (which I'm doubtful of).

It is used a lot, for example vscode finds about 800 lines like this:

if (p_path.begins_with("res://")) {

I didn't find any uses of something.begins_with(""), but there could still be uses of something.begins_with(p) where p points to empty string. But I guess probability of something breaking is very low, although not zero.

[Mickeon]

Through sheer boredom I took a closer look and there is literally not a single instance of an array of chars being passed around that is not a constant. It's likely why this went unnoticed, too.

[akien-mga]

Thanks!

[akien-mga]

Cherry-picked for 4.2.2.