modify helper functions for 'perfect_integration'

[LuLeom]

Describe the bug
Currently, get_obs_var_for_integrated and get_obs_for_integrated in utils/helper_functions.py do not handle well the 'perfect integration' control method. When fetching annotation for the 'perfect integration case, the validation annotations perfectly match the integrated annotations. However, the batch variable remains the same between integrated and validation, potentially raising issues when going through metrics component.
(e.g. fitting a linear model for metrics/average_batch_r2 is not possible as there will only be 1 possible value for the batch variable)

Expected behavior
The helper functions should change the batch number when method == 'perfect integration' so that the batch annotation is swapped (e.g. if a row is annotated as 'batch' == 1, it should be swapped to 'batch' == 2).
Note: ideally the swapping should work in the case of 3+ batches too.

Additional context
PR #36 now contains an exception for 'perfect_integration' where it changes the batch number (hardcoded). The exception must be removed once this issue is solved

[LuLeom]

@ghar1821 We did not act on this issue in the end, right?

I was wondering if solving this issue will actually result in a break somewhere else

[ghar1821]

Nope. I don't think I fixed it when I split the perfect_integration to vertical and horizontal. I remember having to slightly modify these lines of code in metrics/average_batch_r2 when I introduced the perfect_integration_vertical:

if "perfect_integration" in input_integrated.uns["method_id"]:
    input_integrated.obs["batch"] = input_integrated.obs["batch"] + 3

I think with perfect_integration_horizontal, we can just swap out the batch value (so 1 becomes 2 and vice versa). But with vertical, we need to work out which sample we will allocate to batch 2 as all of them comes from batch 1. Though, I don't think it matters which samples we choose. We can just allocate the samples from validation as batch 2 and from unintegrated as batch 1.