Fix and document tests

[necto]

After the change in how lsp-sonarlint fetches the SonarLint executable, tests started to fail on CI, because the SonarLint executable is not downloaded automatically and lsp-mode skips it with only a warning

The following servers support current file but do not have automatic installation: sonarlint

This change introduces an explicit SonarLint download step before running the tests.
I decided to make it an explicit step instead of integrating it into the test harness because it can take several minutes.

Additionally, this PR documents how you can run tests locally, and makes lsp-sonarlint-display-rule-descr-test more reliable.

[necto]

There are 3 kinds of failures remaining:

1. windows-latest + emacs snapshot
   This seems to be Eask problem - it crashes on the emacs (read-from-whole-string ARG) function.
   It might be that Emacs or Eask is improperly installed on Windows.
   I have no Windows machine and little Windows skill to investigate.
   I did manage to hit similar bug when I was trying to install Emacs snapshot on my linux machine with

   guix shell -K --without-tests=emacs-pgtk --with-git-url=emacs-pgtk=https://github.com/emacs-mirror/emacs --with-commit=emacs-pgtk=ce56f939affe8147f4172aa030058029a3c7922b emacs-pgtk
   

The installation command failed some tests (despite --without-tests flag), but I was able to get the Emacs 31.0.5 binary.
When I put it into PATH and tried to run eask status, I got the same error:

Can't read whole string

Note that this job was already failing before this change.
I will remove it from CI for now, and let someone with Windows capability resuscitate it when needed.

2. MacOS/Ubuntu + Emacs snapshot
   The tests run ad infinum printing "Timeout while waiting for response. Method: shutdown".
   In a normal run, this error is printed only a couple of times per test (that's the way we shutdown SonarLint LSP server).
   It seems something is odd with the snapshot + it does not play well with GitHub Actions.
   I was able to reproduce the crazy error-repetition locally using act
   However, unlike GH Action, the tests printed a lot of copies of this same error message and then succeeded in regular time.
   When I attached to the docker container produced by act I was able to run the tests manually
   And I've seen many copies of the same error message.
   However, when ran interactively from Emacs within that container,
   I only see normal amount of these messages in the *Messages* buffer
   I will remove testing Emacs snapshot versions for now.

3. MacOS all Emacs versions
   Here lsp-sonarlint-c++-reports-issues fails (times out). I don't know what is going on here (maybe an issue with path to compiler_commands.json?). @mjburling apparently succeeded at using C++ analyzer, so I would appreciate if you, @mjburling could run the test suite (as explained in the readme version in this PR) and report if it runs well on your machine.

[jcs090218]

1. windows-latest + emacs snapshot

You can ignore this error for now, see emacs-eask/cli#224.

[jcs090218]

We can just remove the windows-latest test.

[necto]

removed

[jcs090218]

Sorry for the confusion. I mean comment out only the windows snapshot.

[necto]

The other two are problematic as well. as you can see, they take all the time in the world (6h), even though they will eventually pass.

[jcs090218]

Ah, okay. I guess we can just comment them all out.

[necto]

done 33f07a2

[jcs090218]

Can we ignore specific tests instead of entirely removing the macos-latest? 🤔

[necto]

Indeed, that's better. I ignored just lsp-sonarlint-c++-reports-issues on macos

[necto]

The tests fail now due to a problem with setup-emacs GH Action. I cannot rerun them to see if the problem is fixed.

[jcs090218]

LGTM. Thank you for your effort on the CI tests! :D