Fix tests: write runtime db

[tkellogg]

Tests broke in #66 but we didn't catch it because our CI/CD pipeline
doesn't run cargo test. This also enables tests on all platforms.

Ping @JakeStanger

closes #72

[JakeStanger]

Ah whoops, good spot.

How come the tests are succeeding on MacOS but not Linux/Windows? Is there a flaky test in there or is it actually something platform-specific?

[dswij]

I have been experiencing this on my machines. I thought it was something wrong on my side.

[tkellogg]

Seems to be 2 problems.

1. During tests, git commands would exit before printing anything if they were unsuccessful. This hid the git commit failure
2. git commit fails on Linux & Windows if user.name and user.email aren't set

[tkellogg]

One more thing: Some places weren't calling .unwrap() on GitRepo::git calls. I'm adding those so that we terminate early automatically. An even better fix would be to make git return a Result so that rustc would warn us if we didn't unwrap.

[tkellogg]

Remaining failure is a race condition. We use Thread::sleep(6), which seems to be not quite long enough sometimes. The fix is to wait only until dura serve starts producing lines of output. I'm not going to fix that yet. I don't have enough time right now. Later.