feat(NODE-4385): add cmap pool pausing functionality

[dariakp]

Description

NODE-4385 / NODE-2994

What is changing?

CMAP implementation updates

• implementation of ready() method
• implementation of PoolClearedError
  • There is follow up work noted in two tickets:
    • one to propagate the original error, which requires sdam-awareness
    • another to make sure the error is fully retryable, which involves a test from a separate DRIVERS ticket
      • this is partially, and possibly fully, addressed in this PR, but I left the TODO ticket in order to verify with the follow up retryability testing
• checkout failure is now exclusively handled by processWaitQueue, as is clearing the wait queue
  • This consolidation was done so that pool paused and pool closed checkout errors could be propagated uniformly to all connection checkout events regardless of whether they were already in the wait queue when the clear or close was called or happened afterwards
• createConnection now always calls the provided callback so that we make sure members on top of the waitqueue awaiting connection establishment don't get left in limbo when the pool is cleared or closed
• pending connections are now exclusively tracked inside the createConnection method for easier handling
• clearing a closed pool should be a no-op

SDAM implementation updates

• Server description transition to data-bearing or direct connection marks pool as ready
• New ResetPool error label is used to clear the pool on server update, when applicable, in order to synchronize pool clearing logic with server marked unknown (per the spec requirement) - this eliminates potential race conditions where a server with a paused pool is still selectable
• Removes the resetConnectionPool event emission because the pool reset is now handled synchronously with server reset update
• Removes duplicate call to resetServer that is already performed by the original failure handler
• Server/topology errors can be any errors: updated logic to wrap incoming error in MongoError so that we can safely check for labels; the original error is stored in a cause property, consistent with the new cause standard on errors in general (available in Node >16.9, so implemented manually for now, for compatibility reasons)

Test updates

• CMAP
  • Unskipping all CMAP tests
• SDAM
  • New prose test
  • Unskipping minPoolsize-error automated (unified) test
• Unified valid-pass
  • Unskipping command tests that are working now

Is there new documentation needed for these changes?

Possibly, depends on what we currently say about the inner workings of the connection pool

What is the motivation for this change?

Avoiding connection storms

Double check the following

• Ran npm run check:lint script
• Self-review completed using the steps outlined here
• PR title follows the correct format: <type>(NODE-xxxx)<!>: <description>
• Changes are covered by tests
• New TODOs have a related JIRA ticket

[dariakp]

I still need to re-review everything here and update the PR description, but it's ready for a primary review