[native] Improvements to the internal timing infrastructure

[grendello]

• Make timing infrastructure shared between MonoVM and CoreCLR hosts.
• Use C++ APIs instead of some of our own ones (std::chrono)
• Make it possible to dump timings into a file instead of logcat
• Timing events no longer require declaring a variable to hold the timing
  sequence ID. Instead, calls to start and end events are now using a thread-local
  stack.
• Add a way to time a single function call by wrapping it in a time_call timing function.

[grendello]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[grendello]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[grendello]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[grendello]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[jonpryor]

Ongoing question, which shouldn't be addressed here, is "should CoreCLR and MonoVM use the same set of properties?"

For example, b217dca parses a debug.dotnet.log system property instead of debug.mono.log, specifically because NativeAOT does not support the "full" set of values & semantics that debug.mono.log provides.

Perhaps "timing" is identical, but this is something to (re)consider in the future.