Update synchronous crash reporting logic

[fractalwrench]

Goal

Updates the existing synchronous reporting of launch crashes so that they fit the behaviour of the new spec. Specifically, synchronous delivery should only be attempted for the most recent startup crash.

Changeset

• "_startupcrash" is appended to the filename based on whether app.isLaunching is set to true, rather than the previous app.duration exceeding a threshold
• Refactored flushOnLaunch() so that it creates a Future and blocks for 2 seconds calling get(). This avoids a busy wait that the previous implementation used
• Altered EventStore so that it only looks for the most recent startupcrash, as determined by the timestamp and "startupcrash suffix encoded in the filename
• Fixed a bug where .json was encoded in the filename twice (this does not affect any functionality)
• flushOnLaunch() no longer calls flushAsync() to improve testability

Testing

• Updated existing test coverage where necessary
• Added new unit tests for logic around finding a launch crash
• Created tests that validate what thread event delivery occurs on, and confirms that flushOnLaunch() blocks the main thread for a short duration

[kattrali]

Looks good, added a few questions/small improvements.