[compiler-rt] Implement __mulbitint3 and __divmodbitint4

[nikic]

We should implement these for parity with libgcc, and to allow LLVM to lower to efficient builtins for multiplication and division of large _BitInts.

See https://gcc.gnu.org/onlinedocs/gccint/Integer-library-routines.html#Bit-precise-integer-arithmetic-functions for the libgcc docs.

[CGDogan]

and to allow LLVM to lower to efficient builtins for multiplication and division of large _BitInts

When this is implemented, it should be ensured that

_BitInt(2560) x,y,z,t;

void process(void) {
z = x/y;
t = x%y;
}

results in a single call to __divmodbitint4