Optimize add and sub

[chfast]

This rewrites the add_with_carry() for uint128 to a bit more efficient form (easier to consume by compilers).
Also subtraction is implemented in the analogous way to additions, instead of reusing addition.

GCC9:

Comparing test/intx-bench-master to test/intx-bench
Benchmark                           Time             CPU      Time Old      Time New       CPU Old       CPU New
----------------------------------------------------------------------------------------------------------------
binary_op256<add>                +0.0007         +0.0007          7186          7191          7186          7191
binary_op256<inline_add>         -0.1695         -0.1695          2317          1924          2317          1924
binary_op256<sub>                -0.1624         -0.1625          8707          7293          8707          7292
binary_op256<inline_sub>         -0.4819         -0.4818          4041          2094          4041          2094
binary_op512<inline_add>         -0.2374         -0.2376          6874          5242          6874          5241
binary_op512<add>                -0.1954         -0.1954          9001          7242          9001          7242
binary_op512<inline_sub>         -0.4907         -0.4907         10430          5312         10430          5312
binary_op512<sub>                -0.1773         -0.1774         12260         10086         12259         10084

Clang9:

Comparing test/intx-bench-master to test/intx-bench
Benchmark                           Time             CPU      Time Old      Time New       CPU Old       CPU New
----------------------------------------------------------------------------------------------------------------
binary_op256<add>                -0.0524         -0.0503          7908          7493          7890          7493
binary_op256<inline_add>         -0.0452         -0.0431          1973          1884          1969          1884
binary_op256<sub>                -0.1284         -0.1258          8552          7454          8527          7454
binary_op256<inline_sub>         -0.3132         -0.3114          2783          1911          2776          1911
binary_op512<inline_add>         -0.5603         -0.5585          9824          4319          9783          4319
binary_op512<add>                -0.0133         -0.0108         11674         11519         11644         11519
binary_op512<inline_sub>         -0.3870         -0.3880          7972          4886          7965          4874
binary_op512<sub>                -0.0672         -0.0692         11794         11001         11794         10978

[codecov-io]

Codecov Report

Merging #115 into master will increase coverage by 2.09%.
The diff coverage is 46.66%.

@@            Coverage Diff            @@
##           master    #115      +/-   ##
=========================================
+ Coverage    64.3%   66.4%   +2.09%     
=========================================
  Files          11      11              
  Lines        1496    1518      +22     
  Branches      166     167       +1     
=========================================
+ Hits          962    1008      +46     
+ Misses        482     458      -24     
  Partials       52      52