NODE-4194 Description
In our Node.js app, after updating MongoDB Node.js driver from v3.6.10 to v4.3.1 we noticed up to 50% performance regression of endpoints that fetch data from MongoDB (v5.0). Each document in the affected collections have at least one field of ObjectId type. After an investigation it looks like parsing ObjectId type fields now takes significant amount of time in comparison with the previous driver v3.6.10.
I came up with the load test that you can run to reproduce the regression:

https://github.com/baryshok/mongodb-object-id-driver-v3

https://github.com/baryshok/mongodb-object-id-driver-v4

It populates `test` collection with 50k docs of shape `

` and fetches all those documents 100 times concurrently.

With driver v3.6 this test takes ~10 secs on my local env, with v4.3 – ~15 secs:

If I switch to using `string` type instead of `ObjectId` for `_id` field: `

` the test results for driver v4.3 become almost equal to the ones for v3.6

The CPU profile for v4.3 shows that ObjectId -> ensureBuffer takes ~3 seconds out of 15 seconds that the whole test takes

I wonder if there's a way to bring back the performance of v3.6 to v4.3 in terms of parsing ObjectId type fields. Thank you!