Fix coverage

[MarcoRossignoli]

Test PR follow up coverlet-coverage/coverlet#700

This is a sample, it's not intended for merge.

I did some debugging and maybe found the issue(we'll see on this PR)

@antonfirsov PR add some out of process tests that slow down a bit the shutdown of test process, you're using msbuild that is affected by know issue https://github.com/tonerdo/coverlet/blob/master/Documentation/KnowIssues.md#1-vstest-stops-process-execution-earlydotnet-test

So new out of process tests create/update hits file(file where we put line/branch hit) on process exits but all other tests fail to log hits because vs test kill test process(the parent process of all other out of ones) before flush microsoft/vstest#1900 (comment)

This PR change the driver to collectors

BTW vstest has got a bug that I've fixed in microsoft/vstest#2221 but not yet released so you need for now to use the preview test sdk.

My local result

cc: @JimBobSquarePants

[claassistantio]

All committers have signed the CLA.

[MarcoRossignoli]

@JimBobSquarePants got and error on codecov upload

2020-01-25T14:11:49.1313263Z ==> Scanning for reports
2020-01-25T14:11:49.5779598Z     + d:\a\ImageSharp\ImageSharp\tests\ImageSharp.Tests\TestResults\6d6166d5-6a42-46eb-93c3-6f4b736c89aa\coverage.info
2020-01-25T14:11:49.5787040Z     + d:\a\ImageSharp\ImageSharp\artifacts\bin\tests\ImageSharp.Tests.ProfilingSandbox\Release\netcoreapp3.1\win7-x64\Microsoft.VisualStudio.CodeCoverage.Shim.dll
2020-01-25T14:11:49.5793378Z     + d:\a\ImageSharp\ImageSharp\artifacts\bin\tests\ImageSharp.Tests\Release\netcoreapp3.1\Microsoft.VisualStudio.CodeCoverage.Shim.dll
2020-01-25T14:11:49.5793911Z ==> Uploading reports
2020-01-25T14:11:49.7850673Z     HTTP 400
2020-01-25T14:11:49.7851344Z Please provide the repository token to upload reports via `-t :repository-token`

[codecov]

Codecov Report

Merging #1094 into master will increase coverage by 16.22%.
The diff coverage is n/a.

@@             Coverage Diff             @@
##           master    #1094       +/-   ##
===========================================
+ Coverage   65.21%   81.43%   +16.22%     
===========================================
  Files         704      704               
  Lines       29311    29311               
  Branches     3290     3290               
===========================================
+ Hits        19115    23870     +4755     
+ Misses       9582     4748     -4834     
- Partials      614      693       +79

Flag	Coverage Δ
#unittests	81.43% <ø> (+16.22%)	⬆️

Impacted Files	Coverage Δ
...c/ImageSharp/ColorSpaces/Companding/LCompanding.cs	0% <0%> (ø)	⬆️
...geSharp/ColorSpaces/Companding/Rec709Companding.cs	0% <0%> (ø)	⬆️
...eSharp/ColorSpaces/Companding/Rec2020Companding.cs	0% <0%> (ø)	⬆️
...mats/Jpeg/Components/Decoder/HuffmanScanDecoder.cs	95.85% <0%> (+2.54%)	⬆️
...Sharp/Formats/Jpeg/Components/Decoder/JpegFrame.cs	95.83% <0%> (+8.33%)	⬆️
src/ImageSharp/Formats/Jpeg/JpegFormat.cs	87.5% <0%> (+12.5%)	⬆️
...p/Formats/Jpeg/Components/Decoder/JpegComponent.cs	100% <0%> (+16.66%)	⬆️
...Formats/Jpeg/Components/Decoder/ProfileResolver.cs	100% <0%> (+16.66%)	⬆️
...arp/Formats/Jpeg/Components/Decoder/AdobeMarker.cs	96.42% <0%> (+17.85%)	⬆️
...lorConverters/JpegColorConverter.FromYCbCrBasic.cs	100% <0%> (+21.05%)	⬆️
... and 114 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 c50eb15...21440b0. Read the comment docs.

[JimBobSquarePants]

@MarcoRossignoli That's fantastic thank you!

This is a sample, it's not intended for merge.

I can't see any reason why we cannot merge this directly. Is there more you intend to add?

I'm actually going to go ahead and merge this. I'm keen to push on and the fix looks great. Thanks again!

[MarcoRossignoli]

I can't see any reason why we cannot merge this directly. Is there more you intend to add?

Didn't know if was your workflow idea. BTW good!

This issue is the reason for us to move most user as possible to collectors, this bug is hard to discover and is related to very "small" behaviour changes.