This adds an optional Scheduler option to source configs. It still
defaults to DFS, our current scheduler. However, for cases like GCP,
alternative scheduling algorithms may be beneficial for large
deployments (as explained in
cloudquery/cloudquery#5389). For this reason,
this change also introduces a new (currently opt-in only) scheduler
called `round-robin`.

A new benchmark scenario has been added to showcase when round robin
really shines. In this scenario, strict global rate limits mean that
waiting for each table to finish all of its clients is not the most
efficient schedule. Instead, a schedule that interleaves the clients for
all tables (as round robin does) performs near optimally, and the
benchmarks show this difference:

```
BenchmarkDefaultConcurrencyDFS
BenchmarkDefaultConcurrencyDFS-8              	       1	     10835 resources/s	73788592 B/op	 1057929 allocs/op
BenchmarkDefaultConcurrencyRoundRobin
BenchmarkDefaultConcurrencyRoundRobin-8       	       2	     11276 resources/s	72121956 B/op	 1054439 allocs/op
BenchmarkTablesWithChildrenDFS
BenchmarkTablesWithChildrenDFS-8              	       1	     31900 resources/s	465282072 B/op	 6859524 allocs/op
BenchmarkTablesWithChildrenRoundRobin
BenchmarkTablesWithChildrenRoundRobin-8       	       1	     33276 resources/s	463475144 B/op	 6856392 allocs/op
BenchmarkTablesWithRateLimitingDFS
BenchmarkTablesWithRateLimitingDFS-8          	       1	        28.30 resources/s	 9098456 B/op	  132129 allocs/op
BenchmarkTablesWithRateLimitingRoundRobin
BenchmarkTablesWithRateLimitingRoundRobin-8   	       1	       821.5 resources/s	 8213856 B/op	  112748 allocs/op
PASS
```

In the `BenchmarkTablesWithRateLimiting` case, Round Robin completes in
just over 1s, DFS completes in 35s.

## Notes for Reviewers

- No functional changes to the DFS scheduler were made in this PR, but
some functions were moved to a different file so they can be re-used by
Round Robin.
- I'm also removing `targetResources/s` from the benchmarks in this PR.
They were useful once, but with rate limiting scenarios it's becoming
too difficult to maintain accuracy.

Follow-up on transformation. Adding unwrap option as needed in some cases where we create our own wrapper struct.

This also moves the tests from codegen.

🤖 I have created a release *beep* *boop*
---


## [1.19.0](v1.18.0...v1.19.0) (2023-01-05)


### Features

* Add scheduler option and introduce Round Robin scheduler ([#545](#545)) ([d89a911](d89a911))
* Add unwrap option to transformations ([#573](#573)) ([a17ee4b](a17ee4b))

---
This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please).