modutime.c: Ensure proper smallint values on overflow

[robert-hh]

This fixes ensure, that the return values of ticks_ms(), ticks_us() and ticks_cpu() fit into a MP_SMALL_INT. That way, they can be fed into ticks_diff() without the risk of an OverflowError (see issue #113). As a side effect, the result of ticks_diff(a, b) is always defined as a reasonable value even if a is a time after b (a > b). In that case, the value is negative. The implementation just copies the mechanism from py/extmod/mphal_utime.c.
Second change: Add the method ticks_add().

[robert-hh]

Letting ticks_ms() and ticks_us() return 64 bit values would not change that behavior, but then ticks_diff() would be obsolete, because if can re replaced by calculating the difference in Python. The code for that could be:

STATIC mp_obj_t time_ticks_ms(void) {
    return mp_obj_new_int_from_ll(system_get_rtc_time() / 1000);
}
STATIC MP_DEFINE_CONST_FUN_OBJ_0(time_ticks_ms_obj, time_ticks_ms);

STATIC mp_obj_t time_ticks_us(void) {
    return mp_obj_new_int_from_ll(system_get_rtc_time());
}
STATIC MP_DEFINE_CONST_FUN_OBJ_0(time_ticks_us_obj, time_ticks_us);

Edit: Changing in addition ticks_diff into:

STATIC mp_obj_t time_ticks_diff(mp_obj_t start_in, mp_obj_t end_in) {
    int32_t start = mp_obj_get_int_truncated(start_in);
    int32_t end = mp_obj_get_int_truncated(end_in);
    return mp_obj_new_int((end - start));
}
STATIC MP_DEFINE_CONST_FUN_OBJ_2(time_ticks_diff_obj, time_ticks_diff);

avoids the overflow exception and gives a reasonable behavior for diffs <= 2**31, and maybe it's the only change needed.