Cache expensive Go compilation and linting

[dgageot]

This will significantly reduce "unit tests" and "checks" jobs on TravisCI.

Signed-off-by: David Gageot david@gageot.net

[codecov]

Codecov Report

Merging #3341 into master will not change coverage.
The diff coverage is n/a.

[tejal29]

I dont' like the idea of using cache for unit test running.
In my past team, we have seen some cases for pants test where some ci runs were false positive due to them being not invalidated.

But again, our team wrote pants cache, and developers write bugs :)
How is your confidence in go tests cache ?

I am ok with trying this and then reverting in case we see false positives :)

[dgageot]

@tejal29 On linux, the test cache is disabled (with --count=1). So I thought the risk would only be that a mac/win test could be flaky and we wouldn't detect that.