Fix arbitrary precision rationals with negative denominator

[psavouli]

This fixes #27.

[codecov]

Codecov Report

Merging #41 into develop will decrease coverage by 3.19%.
The diff coverage is 0%.

@@            Coverage Diff             @@
##           develop      #41     +/-   ##
==========================================
- Coverage    70.11%   66.91%   -3.2%     
==========================================
  Files            1        1             
  Lines          271      272      +1     
  Branches        90       94      +4     
==========================================
- Hits           190      182      -8     
  Misses           3        3             
- Partials        78       87      +9

Impacted Files	Coverage Δ
include/boost/rational.hpp	66.91% <0%> (-3.2%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 47c1b26...ae740e4. Read the comment docs.

[psavouli]

Could I have some help here please? I do not quite know how the coverage report is setup to reproduce it. Equally the master and develop branches show 70% and 67% coverage respectively, even though both branches are the same. So I can't quite tell if there is actually something wrong with my change or something else is going on. Thanks.

[bjodah]

Just wanted to report that this patch works wrt. this issue in downstream SymEngine:
symengine/symengine#1639

[jeking3]

I had to fix the test jamfile to depend on multiprecision for rational_test.cpp, and rebased on develop.

[jeking3]

Coverage report text here is stale (but the heatmap is accurate: 100%), clicking through shows the new code paths are tested.

[jeking3]

Didn't think it would be this messy!