Improve Python unit tests

[matz-e]

Many of the unit tests follow a pattern of:

1. manually populate an index,
2. hard-code a ground truth,
3. perform a query using one particular API.

The idea would be to refactor the tests such that we keep Step 1 and Step 2; but Step 3 we split into a separate function. This function is given the index and the expected ground truth as arguments. The point of this change is that in that function we could check all API variants, instead of only one of them; without creating a lot of repetition.