GODRIVER-2586 Add log messages to CMAP spec

[prestonvasquez]

GODRIVER-2586

Summary

This PR adds log messages to CMAP operations.

Additional Updates

1. Add Reason type to the "event" and "logging" packages

There is an unexported function in the x "topology" package used to source the event reason for the event monitor called "connectionPerished". The exact strings for the event reason differ between the "event" and "logging" packages. In order to unify the two packages, this PR proposes that we update the function signature to return a struct with fields for various reasons:

type reason struct {
	loggerConn logger.Reason
	event      event.Reason
}

2. Unify Component Keys

Add constants to the logger/component.go file that represent "key names". This will unify the IO Sink, the serialization methods, and the key/values appended to logs messages through pool and command operations.

3. Closing v. Checkout Failure

The Go Driver creates it’s pool of connections in the background and there is always a realistic chance that a connection will close after a checkout fails. After careful consideration, the team doesn’t believe this behavior is a bug but a consequence of gracefully trying to concurrently deliver checkouts.

These issues will almost certainly come up again in the specifications PR #1369.

Multiple solutions have been proposed to resolve this issue:

1. Relaxing the requirements on the number of logs, suggested by specifications PR #1372. This was rejected on the basis that we lose the ability to ensure drivers are not generating extra, incorrect log message
2. Guarding the "ready" channel with a mutex in the Go Driver, suggested by PR Guard wantConn.ready with a Mutex #1172. Despite only synchronizing error handling, there is concern from the Go Driver team that this code is confusing.
3. Adding an "unordered" boolean to the “expectedLogMessages.message” objects to indicate that the order for a particular message is not strict. While this has not been rejected, the spec owner wants to discuss the implications of 1369 first.

To prevent GODRIVER-2586 from being blocked, this PR suggests that for the moment we relax the order requirements for logs with "ConnectionClosed" and "ConnectionCheckoutFailed" messages.

Background & Motivation

To add log messages and corresponding tests for connection pool logging.

[matthewdale]

Some comments about synchronization/buffering of json.Encoder, but everything else looks great!

[matthewdale]

One suggestion to fix the ExampleClientOptions_SetLoggerOptions_customLogger test failures. Otherwise, looks good 👍

[qingyang-hu]

Great job!