GODRIVER-3455 Make BSON benchmarks more representative of real use cases.

[matthewdale]

GODRIVER-3455

Summary

• Make BenchmarkMarshal and BenchmarkUnmarshal BSON benchmarks parallel.
• Add more test cases to BenchmarkMarshal and BenchmarkUnmarshal.
• Make it easier to run encoding-specific benchmarks (e.g. BSON, ExtJSON, etc) by restructuring the subtests to make the encoding the top-level subtest.

Background & Motivation

[mongodb-drivers-pr-bot]

API Change Report

No changes found!

[prestonvasquez]

This case doesn't seem like it will be any different than encodetestInstance. But if we keep it, suggest defining it globally and adding it to BenchmarkUnmarshal as well.

[matthewdale]

The marshaled BSON document is the same, but the input type is bson.D instead of a user-defined struct, so the marshal logic is significantly different (e.g. iterating over slice elements vs struct fields).

[prestonvasquez]

My comment was unclear, I'm suggesting we do something like this:

var encodetestBsonD bson.D

func init() {
	b, err := bson.Marshal(encodetestInstance)
	if err != nil {
		log.Fatalf("error marshling struct: %v", err)
	}

	err = bson.Unmarshal(b, &encodetestBsonD)
	if err != nil {
		log.Fatalf("error unmarshaling BSON: %v", err)
	}
}

[prestonvasquez]

LGTM with consideration for this comment: https://github.com/mongodb/mongo-go-driver/pull/1974/files#r2000139379