Tables interface for FMUSolution

[AnHeuermann]

Changes

• Implementing Tables.jl interface
  • Column access defined
• Removed DataFrames extension, it's handled by Tables.
• Simplified fmiSaveSolutionCSV, CSV.write used Tables interface as well.

Related Issue

Fixes #198

[AnHeuermann]

Creating DataFrames from solution::DiffEqCallbacks.SavedValues{Float64, NTuple{76, Float64}} works fine, but I'm not sure why it breaks fmiLoadSolution for JLD2.

[ThummeTo]

Maybe a general issue with the latest JLD2 release... this issue is not limited to your PR

[AnHeuermann]

Seems you are right, using

[compat]
JLD2 = "< 0.4.37"

works locally.

[AnHeuermann]

I noticed an issue with my implementation in Tables.jl. It is wrong to add all alias variables to the column names. We should only add the ones that where in recordValues of the fmiSimulate call.
@ThummeTo is the array recordValues stored in the simulation result?

[ThummeTo]

It is!

[ThummeTo]

The JLD2-thing seems resolved ...

[AnHeuermann]

I'll see if I can finish this PR soonish.

[AnHeuermann]

It is!

I can't find the list of variable names recordValues inside struct FMU2Solution or FMU3Solution. The list of value references is stored in valueReferences, but this is not unique.
The case I'm having trouble with is when FMU A.fmu has alias variables a=b. Both will have have the same value reference, but if I simulate with

solution = fmiSimulateME(fmu, (0., 1.); recordValues=["a"])

only a should be in the list of names of the table.

If that is not available we can keep it as it is now: Store both a and b in table.