fix(postgrest): race condition when setting fetchOptions and execute method call

[grdsdev]

What kind of change does this PR introduce?

Fixes a bug which could lead to a race condition when executing the request on Postgrest.

What is the current behavior?

Lock was being acquired first for updating the fetchOptions and then inside the execute method for executing the request, this could lead to the execute method to use the wrong fetch options.

What is the new behavior?

Move assignment of fetchOptions inside execute method, on a single lock acquisition.

[steve-chavez]

Lock was being acquired first for updating the fetchOptions and then inside the execute method for executing the request, this could lead to the execute method to use the wrong fetch options.

@grdsdev Q: Why are locks needed on the client? Can't it be stateless? (Since PostgREST requests are stateless)

[grdsdev]

@steve-chavez

Check the following test:

func testRaceCondition() async {
    let select = client.from("users").select()

    Task {
      print("Task 1")
      try await select.execute(options: FetchOptions(head: true))
    }

    Task {
      print("Task 2")
      try await select.execute(options: FetchOptions(head: false))
    }

    try? await Task.sleep(nanoseconds: NSEC_PER_SEC * 2)
  }

Task 1 and Task 2 can run on different threads, Task 2 can even start running before Task 1, that is all up to the Swift concurrency system.

[steve-chavez]

@grdsdev Got it. Thanks for the explanation 👍