Only download chrome driver

[pranavkm]

This should avoid extra web traffic by only installing the chrome driver. Inspired by a build failure due to transient network issue:

2020-01-28T21:33:29.0255009Z YARN : error : Could not download https://github.com/mozilla/geckodriver/releases/download/v0.23.0/geckodriver-v0.23.0-win64.zip [F:\workspace\_work\1\s\src\Components\test\E2ETest\Microsoft.AspNetCore.Components.E2ETests.csproj]
2020-01-28T21:33:29.0257959Z F:\workspace\_work\1\s\src\Shared\E2ETesting\E2ETesting.targets(23,5): error : Command failed with exit code 1. [F:\workspace\_work\1\s\src\Components\test\E2ETest\Microsoft.AspNetCore.Components.E2ETests.csproj]
2020-01-28T21:33:29.2096369Z ##[error]Build failed.

[SteveSandersonMS]

It's unusual to see Windows-style paths in a package.json file. Does it work with / too? If so that would be preferable as we may want this to run on other OSes too.

[SteveSandersonMS]

If we're hardcoding the seleniumConfigPath into package.json anyway, why go to all this trouble to pass it dynamically here? Would it be simpler to modify the selenium-standalone script in package.json so that it also hardcodes the config path? Then we wouldn't need to pass it from here, and therefore wouldn't need the AssemblyMetadataAttribute, and wouldn't need the change to E2ETesting.props, etc.

[pranavkm]

I think the issue is that knowing the relative path inside of a test is a bit tricky. Using relative paths can mess up depending on how the test is launched and it's not always correct to assume PWD is the test directory. In other places where we need to find something relative, we do something quite similar - have the build tell us where things are.

[SteveSandersonMS]

OK, sounds like a good reason!