refactor(tests): Replace MockLanguageServer singletons with configurable, per-testcase LanguageServer instances

[FlorianKroiss]

The tests which require interaction with a Language Server use a single MockLanguageServer instance per test, which has a few draw-backs:

1. The MockLanguageServer does not behave like a proper LS. Namely, it's exit() did nothing, where a "real" LS would have to close the outbound connection to the client. However, since the MockLanguageServer serves as the endpoint for many connections, it could not know which one called exit()
2. Tests which start multiple servers cannot track the state of the individually launched servers, because they all are connected to the same instance.
3. Tests had to make sure to extend AbstractTest or use AllCleanExtension so that a new MockLanguageServer is instantiated and no State is leaked between tests.

This PR introduces a MockLanguageServerExtension which enables individual test cases to configure instances of the MockLanguageServer which are tailored towards the needs of the test case.
The MockLanguageServerFactory can be used to do so and is automatically injected into a test parameter of the same type. A dummy example test case might look like this:

	@Test
	public void myTest(MockLanguageServerFactory factory) throws Exception {
		// Executed for each created server.
		factory.withConfiguration((idx, server) -> {
			if (idx == 0) {
				// Server 0 delivers one entry
				server.setCodeLens(List.of(new CodeLens()));
			} else {
				// Server 1 delivers two entries
				server.setCodeLens(List.of(new CodeLens(), new CodeLens()));
			}
		});
		// Override default capabilities for created servers
		factory.withCapabilities(() -> {
			var cap = MockLanguageServer.defaultServerCapabilities();
			cap.setCodeActionProvider(Boolean.TRUE);
			return cap;
		});
		
		// Open file which will launch a server
		IFile file = TestUtils.createUniqueTestFile(project, "someFile");
		ITextViewer viewer = TestUtils.openTextViewer(file);
		// [... do stuff with viewer ...]
		
		// Change configuration of started server
		factory.getServer().setCodeLens(null);

		// Assert stuff on started servers
		assertTrue(factory.getServerCount() == 2);
		assertTrue(factory.getServers().get(1).getState() == MockServerState.RUNNING);
	}

Note that it's not required to call any method on MockLanguageServerFactory, in which case a default configuration will be used. It's not even necessary to add the MockLanguageServerFactory factory parameter.

Most of the code changes in this PR are straight forward, i.e., replacing calls to MockLanguageServer.INSTANCE with the appropriate new mechanisms from MockLanguageServerFactory.

A few more cleanups that are also contained in this PR:

• MockLanguageServerMultiRootFolders/MockConnectionProviderMultiRootFolders and associated entries in the fragment.xml have been removed. These were outdated copies of the MockLanguageServer, with the only difference, that they advertised multi-root support. Tests which require a LS with multi-root support, can now just set the respective capabilities with factory.withCapabilities(MockLanguageServer::multiRootCapabilities)
• AbstractTest and AllCleanExtension have been removed, as their functionality is superseded by MockLanguageServerExtension

[FlorianKroiss]

My original motivation for these changes is that the Mock LS does not properly close its outputstream on server exit.
Fixing this is one of the requirements to fixing all of the numerous AsynchronousCloseException/Broken Pipe/ClosedByInterruptExceptions in our test execution.

@rubenporras Can you please look at a few modified tests and share your opinion on the proposed changes?

Edit: Maybe @sebthom or @mickaelistria also have an opinion on these changes?

[sebthom]

I don't have time right now to do a thorough code review, but I strongly support the direction of this refactoring. I had considered doing something similar myself at one point, so I'm glad to see this happening.

[rubenporras]

Hi @FlorianKroiss ,

I do not have the time to properly look at the PR, but if it addresses the AsynchronousCloseException/Broken Pipe/ClosedByInterruptExceptions then I am very happy if you can finish this.

[FlorianKroiss]

I do not have the time to properly look at the PR, but if it addresses the AsynchronousCloseException/Broken Pipe/ClosedByInterruptExceptions then I am very happy if you can finish this.

It's at least a step in that direction.
From my side, I think this is good to go, as we can still fine-tune it later on.
I think the bot remarks can be dismissed, as the relevant code was only moved.