Issue in arm_div_int64_to_int32(int64_t num, int32_t den): There may be a risk of overflow when the num is 0x800000000000000

[Angerla]

For the above code, if the incoming data is 0x800000000000000, there seems to be a risk of overflow. In some compilers, the data of (int64_t)LONG_MIN may be converted to -2147483648:

When the incoming data is 0x800000000000000 it will go to the processing of -num and the number of -(-9,223,372,036,854,775,808) may be overflow in the range of int64_t (-9,223,372,036,854,775,808~9,223,372,036,854,775,807). And this might lead to the generation of undefined behaviors of the compiler.

[christophe0606]

@Angerla Latest commit should fix the issues.
I have also added a saturation when result is on int64 before converting to int32.
test patterns have been updated.

[christophe0606]

gcc build issue to correct

[Angerla]

Thanks!