Unneeded "weak"ness of "weak_alias"

[nyh]

This is a low-priority "nitpick" kind of issue.

OSv code uses the weak_alias() macro in dozens of places, directly or through the LFS64() macro. I think we should have an alias() macro and use that instead of weak_alias() in many, if not all, the places where weak_alias() is currently used.

weak_alias() appears it used for two reasons:

1. Musl code (libc//.c) defines the real function with a name __mktemp(), and then makes the regular name mktemp() a weak alias for that. That supposedly allows a program to redefine mktemp() to something else, and still use the library function, calling it __mktemp().
2. LFS64() and two or three other uses of weak_alias(). Here, we don't really need a weak symbol. In my opinion, it makes little sense to have stat() a strong symbol which cannot be overriden by the program, and yet stat64() is weak, and can be overridden.

Finally, I don't think we even need weak symbols for #1 above. I believe our linker always look first in the program and not the kernel, so any symbol can be overriden by the program, not just weak symbols. I believe we ignore the "weak" parameter of symbols.