math.big: change how to hanle negatives in div_mod(), add tests

[tankf33der]

Tests against gmp show that div_mod() behaves incorrectly.
Everything should be in truncate format regardless of input parameters.
This patch fixes this issue.

1. Now /, % and div_mod works as mpz_tdiv_q, mpz_tdiv_r and mpz_tdiv_qr respectively. Tested against gmp and flint libraries.
2. Added tests in big_test.v compared against julia lang.

[huly-for-github]

Connected to Huly®: V_0.6-23067

[spytheman]

Excellent work.

[spytheman]

The CI failure seems unrelated, but I'll need to check more thoroughly 🤔 .

[spytheman]

The CI failure was indeed unrelated - after the third retry it worked, and it also happened for another PR.

[tankf33der]

Thank you for merge.

I decided to add a comment for history and science - my function in julia that I used to verify tests is divrem().

Successfully passing 100M loop against gmp's mpz_tdiv_qr gives me more confidence that I implemented everything correctly.