chore: split CI builds to new files

[codebytere]

Follow-up to #455:

I also did a bit of a refactor here, since there were a million classes in the result parser file and it was getting wildly unwieldy. I tried to do the minimum necessary to make this work but I plan to follow it up with the remainder of the the class split-out work.

This is the remainder of that work - the file was inscrutably long and so i felt this would be more maintainable in this form. Specs are passing, but please take a good look if you can to make sure i didn't whiff on anything!

cc @nodejs/node-core-utils

[codecov]

Codecov Report

Merging #464 into master will increase coverage by 5.69%.
The diff coverage is 65.84%.

@@            Coverage Diff             @@
##           master     #464      +/-   ##
==========================================
+ Coverage   76.88%   82.58%   +5.69%     
==========================================
  Files          28       34       +6     
  Lines        1752     1665      -87     
==========================================
+ Hits         1347     1375      +28     
+ Misses        405      290     -115     

Impacted Files	Coverage Δ
lib/ci/ci_utils.js	36.84% <14.28%> (-63.16%)	⬇️
lib/ci/build-types/fanned_build.js	18.42% <18.42%> (ø)
lib/ci/build-types/linter_build.js	26.66% <26.66%> (ø)
lib/links.js	88.73% <50.00%> (-1.13%)	⬇️
lib/ci/build-types/commit_build.js	79.31% <79.31%> (ø)
lib/ci/build-types/normal_build.js	80.95% <80.95%> (ø)
lib/ci/build-types/benchmark_run.js	87.80% <87.80%> (ø)
lib/ci/build-types/test_run.js	88.88% <88.88%> (ø)
lib/ci/build-types/pr_build.js	92.68% <92.68%> (ø)
lib/pr_checker.js	95.49% <100.00%> (ø)
... and 7 more

---

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 48c306b...53bc461. Read the comment docs.

[priyank-p]

Not saying this needs to be done this way, but I think doing this refactor one or couple of class per commit would have made reviewing this way easier. Less changes to look at per commit means it is easier to spot a mistake :)

On the contrary, if you were missing an import of a class then eslint would have complained but doing it that way is still better.

[joyeecheung]

Rubberstamp LGTM :)