pruntime: Structured logging with tracing

[kvinwang]

Log lines for concurrent requests are intermixed making it difficult to trace the logic flow. The PR replaces some logs with the tracing crate so that we get structured log lines that can be far easier to understand.
The log looks like this:

As you can see there is a line id=114 that is mixed between 119 and 120.

How it works:

• The crate tracing is designed for structured logging. It can set stacked spans as the scope of each log printing.
• We use tracing_subscriber::fmt instead of env_logger to init the logger. The tracing_subscriber::fmt is a tracing-subscriber for tracing that prints logs to the console just like env_logger to log. tracing_subscriber::fmt is also a logger for log which bridges the log::* to the tracing framework.
• We set some tracing instrument points:
  • Each Rocket routes entry point. We assign each request an id so that we can clearly see each log line is printed within which RPC request. (We do this because the Rocket framework itself doesn't support tracing so far. There is a draft PR, but got inactive for a long time)
  • The dylib pink runtime entry point. To pass the request id to the tracing framework inside.
  • The dylib pink runtime ocalls point.
  • The sidevm instance serving entry point.

[h4x3rotab]

Very useful and looks promising

[h4x3rotab]

Nice. Also simplified a lot of logging code.

[h4x3rotab]

How about document it somewhere?

[kvinwang]

Didn't find a good place elsewhere. Let me just add some comments to this file.

[h4x3rotab]

Maybe we can create a simple doc file for pruntime startup flags.

[kvinwang]

Added docs/pruntime-flags.md