Cleanly exit Chatterino instead of force exiting

[Mm2PL]

Description

Joining the thread stops us from crashing.

Known issues relating to this:

• if you start and close Chatterino after the recent-messages request was sent but before the response was received, Chatterino will crash. This didn't use to be the case because we'd just throw the process into _exit.

[pajlada]

Needs some more testing, specifically on:

• macOS
• Windows - nerixyz tested this and gave a patch that fixes his issues
• Windows with browser extension - brian6932 tested this and had no issues
• Linux

[pajlada]

Seems like we can't use std::jthread

[Nerixyz]

This currently causes a call to std::terminate on Windows, the following fixed it:

diff --git a/src/providers/liveupdates/BasicPubSubManager.hpp b/src/providers/liveupdates/BasicPubSubManager.hpp
index f82f70363..3a61ee8fe 100644
--- a/src/providers/liveupdates/BasicPubSubManager.hpp
+++ b/src/providers/liveupdates/BasicPubSubManager.hpp
@@ -94,7 +94,10 @@ public:
                 .toStdString());
     }
 
-    virtual ~BasicPubSubManager() = default;
+    virtual ~BasicPubSubManager()
+    {
+        this->stop();
+    }
 
     BasicPubSubManager(const BasicPubSubManager &) = delete;
     BasicPubSubManager(const BasicPubSubManager &&) = delete;

We should make sure the app always eventually exits (not like in the tests some time ago, where the Qt event-loop wouldn't spin). I wanted to add the option to restart Chatterino on a crash back to Windows soon, so we should also ensure that we don't abort/terminate/crash when exiting.

[brian6932]

Tested closing on Windows with the browser extension attached, and had no issues

[pajlada]

New commit has been tested on i3 & macOS

Only thing I notice is that 7tv's eventapi takes a while to close

[github-actions]

clang-tidy made some suggestions

[pajlada]

Latest commit tested on i3 & macOS - works as expected & shuts down pretty fast

[Nerixyz]

This was broken by #5058 and #5059 will break this on Windows.

[pajlada]

I think I would prefer to do smaller steps towards removing _exit, essentially lean into the fakeDtor we have now and start cleaning up one part of Chatterino at a time, until fakeDtor essentially runs the full destructor

[pajlada]

oops wrong pr :tf:

[pajlada]

Closing in favour of #5537