Fix intermittent test failures

[jung-kim]

Test fails intermittently quite often and considering various combinations of tests we run, test usually never passes.

https://github.com/FredrikNoren/ungit/runs/4005712725?check_suite_focus=true

This maybe simple as doing more retries during tests

[campersau]

I know macos clicktests are very flaky with GitHub Actions. On travis (which we need to migrate to .com) they often pass like here https://travis-ci.org/github/FredrikNoren/ungit/builds/772284376. That's why I have added a retry for clicks in 934d152:

ungit/clicktests/environment.js

Lines 269 to 282 in d612c0f

	async click(selector, clickCount) {
	let elementHandle = await this.waitForElementVisible(selector);
	try {
	await elementHandle.click({ clickCount: clickCount });
	} catch (err1) {
	elementHandle = await this.waitForElementVisible(selector);
	try {
	await elementHandle.click({ clickCount: clickCount }); // try click a second time to reduce test flakiness
	} catch (err2) {
	winston.error(`Failed to click element: ${selector}`);
	throw err2;
	}
	}
	}

But since I don't have a mac these issues are very hard to debug for me.

[jung-kim]

I think number of tests has gotten bit out of control. I don't there is a much value in testing in so many node versions and OS versions. To me, node works on an OS, then it should be enough. This is because if a test fails on specific OS, then it's a library or node issue, not ungit issue.

If we were to have tests within different browsers, then that would be helpful. But not various tests for different OSs.

I think we should cut down on appveyor and travis and stick with github actions and make it stable

[jung-kim]

Closing this issues now. #1505 has more details on current state of click testing.

While there are still some outstanding race conditions, tests has stabilized quite a lot.