fix deprecated indexing in EnsembleSummary plot recipe

[SebastianM-C]

I missed this ones.
@AayushSabharwal is this ok? It seems a bit odd to need sim.u.u[1] at first sight.

[codecov]

Codecov Report

Attention: 4 lines in your changes are missing coverage. Please review.

Comparison is base (5d3ac43) 39.96% compared to head (d6bb6e2) 39.96%.

Files	Patch %	Lines
src/ensemble/ensemble_solutions.jl	0.00%	4 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #604   +/-   ##
=======================================
  Coverage   39.96%   39.96%           
=======================================
  Files          54       54           
  Lines        4076     4076           
=======================================
  Hits         1629     1629           
  Misses       2447     2447           

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

[AayushSabharwal]

Yeah that is weird. VectorOfArrays shouldn't have a VectorOfArray as u. I'll investigate this further

[SebastianM-C]

I'm not sure if it helps, but I was wondering if this is due to the fact that individual ODESolutions are no longer AbstractArrays and the recipes relied on this to distinguish between scalar and vector cases, as for ensemble you can have scalar differential equations or output functions that return scalars.

[AayushSabharwal]

This does seem to be what is required here. I'm not fully on board with the AbstractVectorOfArray.u being another AbstractVectorOfArray thing that's going on, but if it works it works I guess.

[SebastianM-C]

Should this be merged?

cc @ChrisRackauckas