[WIP] Fix Ensemble indexing when output_func is used

[SebastianM-C]

When output_func is used, the resulting type in sim.u[i] is no longer <: AbstractVectorOfArray, so we need to avoid assuming that like in #599, #602 and #604

I was thinking of the following output functions as tests:

function output_func(sol, i)
    (x=sol[1, :], t=sol.t), false
end

function output_func(sol, i)
    last(sol), false
end

function output_func(sol, i)
    [first(sol), last(sol)], false
end

For the plot recipes I was thinking that we could test if no warnings are thrown by the plot calls. That should catch depwarns too and if there's any error in the recipes, the test would error.

This PR is WIP and not yet ready.

[codecov]

Codecov Report

Attention: 5 lines in your changes are missing coverage. Please review.

Comparison is base (e46e491) 39.96% compared to head (4a8e883) 39.94%.

Files	Patch %	Lines
src/ensemble/ensemble_analysis.jl	0.00%	5 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #607      +/-   ##
==========================================
- Coverage   39.96%   39.94%   -0.02%     
==========================================
  Files          54       54              
  Lines        4076     4078       +2     
==========================================
  Hits         1629     1629              
- Misses       2447     2449       +2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[AayushSabharwal]

LGTM. Are these functions tested anywhere, in both cases (.u isa AbstractVectorOfArray and otherwise)?

[SebastianM-C]

There are still some cases to handle. As for tests, the recipes don't have any tests, I was thinking to ad some with @test_nowarn.