Fix suspect crash cause in 5.0.0

[fddecc]

What did you change and why?

A change was introduced between 4.1.5 and 5.0.0 that is resulting in crashes. Multiple internal crash reports (as well as #58) indicate an issue on

TrueTime.swift/Sources/NTPConnection.swift

Line 67 in 228bc5e

lockQueue.sync {

.

Internal reports indicated EXC_BAD_ACCESS. This would be expected to occur in sync block in case self is de-allocated.

When comparing 4.1.5 to 5.0.0 (4.1.5...5.0.0#diff-77256cb830a28637414932596218ae34R125), relocated release of unmanaged self is suspect.

This pr partially reverts c301ea8.

Potential risks introduced?

Even more crashes?

What tests were performed (include steps)?

Additional code was introduced to manually release NTPConnection object (similar to what was done prior to this change) and was able to repro the crash. It goes away after the changes in this pr. But more comprehensive testing might be needed.

Checklist

• Unit/UI tests have been written (if necessary)
• Manually tested

[fddecc]

For bit more context: my thinking is that close() invoked from stopQueue() in NTPClient was what was causing this. It would release NTPConnection object, and then the canRetry lockQueue.sync block would fail since it was referencing self that was deallocated. But bit difficult to determine. Commit where this change was introduced did not have a pr or additional details other than title, so bit hard to gage if we are re-introducing a major issue. However, given that this is fairly close to what's in 4.2.0 and that has been stable for us, it gives me some confidence that this won't be a regression.