This PR makes some small tweaks to code coverage reporting in codecov.

We don't need to provide github.com/coder/coder/codersdk explicitly as coverpkg since we're already passing all packages via ./.... In .vscode/settings.json this was set to ./. which didn't match any package => changed to ./....

When viewing the reports on codecov.io I noticed that some lines that should be covered, were not. I believe this stems from us uploading short coverage reports from the mock database run. It should be sufficient to only upload from the postgres test. The downside is that it's the flakier test, so it might take a while before the coverage can be observed. This is actually most likely due to the postgres test failing so often.

Finally, the *test packages where added to coverage ignore. Although, thanks to -coverpkg=./... they have great coverage, but I don't think it's a meaningful metric for us to include (they're essentially test code in public packages).

Side note: I realized codecov reports lines covered whereas go reports statements, I believe this is a reason our coverage appears worse on codecov than reported by go test -cover.

• fix: Remove explicit coverpkg github.com/coder/coder/codersdk
• fix: Stop uploading short / mock db test coverage
• fix: Ignore test utils in coverage (clitest, coderdtest, ptytest)