optimize AddMod where modulus is 4-limbs wide and x/y highest limb is less than or equal to modulus highest limb

[jwasinger]

uint256 master:

BenchmarkAddMod/small/uint256-4                 71864941                14.78 ns/op
BenchmarkAddMod/mod64/uint256-4                 45851817                23.58 ns/op
BenchmarkAddMod/mod128/uint256-4                23512131                49.20 ns/op
BenchmarkAddMod/mod192/uint256-4                22141264                53.37 ns/op
BenchmarkAddMod/mod256/uint256-4                55604551                21.38 ns/op

this PR:

BenchmarkAddMod/small/uint256-4                 74203911                14.75 ns/op
BenchmarkAddMod/mod64/uint256-4                 44762869                23.61 ns/op
BenchmarkAddMod/mod128/uint256-4                22504081                49.59 ns/op
BenchmarkAddMod/mod192/uint256-4                21490502                52.70 ns/op
BenchmarkAddMod/mod256/uint256-4                65233677                17.39 ns/op

benchmarks ran with turbo-boost disabled so there is little deviation in the results between runs.

[codecov]

Codecov Report

Merging #120 (01d9874) into master (78905e8) will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##            master      #120   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            4         4           
  Lines         1403      1418   +15     
=========================================
+ Hits          1403      1418   +15     

[holiman]

Thanks, sorry for letting it sit!