"Big calculations" sanity-tests

[saxbophone]

Currently most thorough arithmetic testing relies upon randomly-generated inputs. This is fine for thorough and varied coverage, however gigantic values (that exceed uintmax_t on test system) are not used.

I'd like to write some additional separate unit tests that use the custom literals to test certain hardcoded sums on huge numbers, something like exponentiation would be a good idea, but most arithmetic operations should be tested in this way. An online calculator offering full precision in its results should be used to generate the expected return values —something like Wolfram Alpha wouldn't be a bad choice.

[saxbophone]

We better choose our choices of calculations real carefully here, with exponentiation we are likely to run afoul of some limit somewhere for the maximum literal length, as these things reach thousands of digits very very quickly...