new test setup

[shunjizhan]

Context

Previously the test setup was a little awkward, that we didn't run an actual server, but use supertest to "run" the server inside the test. So we had to maintain extra utils to adapt how to call the endpoints. Also, since tests are too coupled with server, it would be very hard to extend for more complex e2e tests.

Change

Decoupled server and tests, and now it's a much cleaner setup: it starts a singleton server as usual, with nyc tracker instrumented. So we can run tests normally, and still get the coverage reports.

This should lay a solid ground for bodhi test v2 too.

Note that this setup only works with nyc, and c8 seemed to have some odd compatibility issue.

Test

whole flow seems to work as expected

test coverage dropped a little, since nyc (now) and c8 (before) has different ways of calculating coverage

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 84.97%. Comparing base (d89ede5) to head (aa56b01).
Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master      #75      +/-   ##
==========================================
- Coverage   88.56%   84.97%   -3.60%     
==========================================
  Files          26       27       +1     
  Lines        1863      619    -1244     
  Branches      164       69      -95     
==========================================
- Hits         1650      526    -1124     
+ Misses        213       91     -122     
- Partials        0        2       +2     

Files with missing lines	Coverage Δ
src/middlewares/router.ts	100.00% <100.00%> (ø)

... and 23 files with indirect coverage changes