OS_CallAfter assumes times are signed

Spotted a while ago, but finally remembered to look it up in the PRM.
The time (in R0) isn’t specified as signed, it just says that “an invalid value” will return an error. Similarly for OS_CallEvery since the two are related.

Internally the kernel uses the seed as a countdown value, ie. it’s not doing maths on it compared with the monotonic time (which could itself wrap while waiting for the call after to elapse).

So I can’t think of a reason not to change it to allow the full 497 days that a 32 bit centisecond would allow since time only ever goes up in this universe.