Fix crash in sentry_toIso8601String

[philipphofmann]

Description

The SDK sometimes crashes with SIGABRT in sentry_toIso8601String.

Exception Type: EXC_CRASH (SIGABRT)

Thread 0 Crashed:
0   libsystem_kernel.dylib          0x1e54a2974         __pthread_kill
1   libsystem_pthread.dylib         0x1f8ee10e8         pthread_kill
2   libsystem_c.dylib               0x1a4f5bb7c         abort
3   libsystem_malloc.dylib          0x1ad186c64         malloc_vreport
4   libsystem_malloc.dylib          0x1ad186f0c         malloc_zone_error
5   libsystem_malloc.dylib          0x1ad17ca40         nanov2_guard_corruption_detected
6   libsystem_malloc.dylib          0x1ad162f80         nanov2_allocate_outlined
7   CoreFoundation                  0x19cfa8ab4         _CFRuntimeCreateInstance
8   CoreFoundation                  0x19cfa8540         __CFStringCreateImmutableFunnel3
9   CoreFoundation                  0x19d01ef04         CFDateFormatterCreateStringWithAbsoluteTime
10  Foundation                      0x19bedaab4         -[NSDateFormatter stringForObjectValue:]
11  Sentry.framework                0x1078f7670         -[NSDate(SentryExtras) sentry_toIso8601String]
12  Sentry.framework                0x1078e6a58         -[SentryBreadcrumb serialize]
13  Sentry.framework                0x107908c48         -[SentryScope addBreadcrumb:]
14  Sentry.framework                0x1078d9a88         -[SentryHub addBreadcrumb:]
15  Sentry.framework                0x107907bf4         +[SentrySDK addBreadcrumb:]

We also see this in our internal SDK crashes.

[philipphofmann]

I double-checked the code and didn't find anything that we can do to avoid this issue. Looking at more stacktraces, it looks like the NSDateFormatter crashes because it can't allocate the memory for the string. I added an extra test in #4983 to ensure the method doesn't crash when passing nil to it.

It seems like the application is in a state where it would crash anyway. Instead, we could avoid such crashes by disabling certain features when the app is under high memory pressure. As we don't see a high volume for these crashes, I'm going to close the issue.

While investigating this issue, I realized we could slightly speed up adding breadcrumbs and opened a PR: #4984.