[9.x] Revert some Carbon::setTestNow() removals

[derekmd]

Revert about half of cleanup PR #40790 that removed "useless" calls to Carbon::setTestNow(). Now some tests are non-deterministic and can intermittently fail for new PRs being submitted:

e.g., https://github.com/laravel/framework/runs/5798854285

1) Illuminate\Tests\Cache\CacheFileStoreTest::testIncrementDoesNotExtendCacheLife
Expectation failed for method name is "put" when invoked 1 time(s)
Parameter 1 for invocation Illuminate\Filesystem\Filesystem::put('D:\a\framework\framework\test...da8a33', '1648892146i:2;', true) does not match expected value.
Failed asserting that two strings are equal.
--- Expected
+++ Actual
@@ @@
-'1648892145i:2;'
+'1648892146i:2;'

The mocked cache expiry was unexpectedly 1 second in the future.

---

• Carbon::setTestNow(Carbon::now()) keeps time-based mocks and assertions deterministic. Otherwise if the clock's second ticks over mid-test, assertions and mockery checks can be run on an unexpected time of day.
• tests/Validation/ValidationValidatorTest.php
  • Avoid putting Carbon::setTestNow(null) at the end of the test case method. If that test fails, Carbon::setTestNow(new Carbon('2018-01-01')) is now leaking into subsequent tests.
  • Worst case, wrap it in a try { /* assertions */ } finally { Carbon::setTestNow(null); }.

[driesvints]

Ping @lucasmichot just FYI ^

[kylekatarnls]

For the record, Carbon 3 will be a bit more strict on this.

Carbon::setTestNow(Carbon::now()); is about freezing the time to make as if the whole test method was running instantly.

Typical case would be things like this:

$repo->getStore()->shouldReceive('put')->once()->with('foo', 'bar', 602);
$result = $repo->remember('foo', Carbon::now()->addMinutes(10)->addSeconds(2), static fn () => 'bar');

(see #41824)

Here the moment to be chosen as the mocked "now" does not matter but it's important that the time is actually frozen during the test.

Between the moment you call Carbon::now() in the test and the moment this date will be read and converted into a number of seconds, few microseconds will elapse. Then difference between this passed DateTime and the new "now" reference will be something like 601.999123 seconds. With Carbon 2 microseconds was brutally truncated, so such a test with no time mock would succeed most of the time and fail if the microsecond unluckily overflow to previous second. In Carbon 3, it will be always take microseconds into account and truncate the result so systematic failure.