Because that main ssh session runs as `/bin/sh`, which doesn't support
`[[  ]]`

If the expected data mismatched, test failed, make was stopping
executing the recipe and nothing was actually updated. This commit fixes
it to ignore failure in `make test`, because failures are expected.

Just to make the output more useful, to figure which log line
corresponds to which test case.

There were the following errors:

```
fatal: detected dubious ownership in repository at '/home/runner/work/nerdlog/nerdlog'
```

So this commit fixes it by makring that directlry as trusted for git.
They weren't actually causing the test run to fail, but still better to
fix it, since it would bite us later.

Just to cover the ssh transport with tests, too.

Because they'll be reused in end-to-end tests as well, which are going
to be in a different package.

So the flags being added are: `--lstreams-config`, `--cmdhistory-file`,
`--queryhistory-file`. It might be useful in general, but the reason
it's implemented now is that it'll help for the end-to-end tests.

These tests cover the actual TUI output, captured by tmux.

This way we can be sure that these exact binaries, that we published as
release assets, actually work.

Just to make things easier to reason about, and to take advantage of
convenient filters like:

```yaml
on:
  push:
    branches:
      - master
      - release-*
```

(This `release-*` thing is not there yet though, but likely will be)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>