Add overflow-checking variants of arithmetic dyn kernels (#2740)

* Init

* More

* More

* Add tests

* Fix clippy

* Remove macro

* Update doc

* Fix clippy

* Remove length check

* Tweak try_binary to coordinate latest optimization

* Fix clippy

* Use for loop

* Split non-null variant into never inline function

* Add value type check

* Multiply by get_byte_width of output type.

Author: viirya