Test for error value from apply_recipe fallback.

[BioTurboNick]

For compatibility with JuliaPlots/Plots.jl#3765

Catching an exception here as part of the normal flow can vastly slow down plotting, as it happens frequently. Returning an error value instead, which the calling function can check, can speed up plotting 2x.

Addresses #93

[t-bltg]

Looks like our CI is broken, I'm trying to fix that.

In the mean time, looks like some failures are related to this PR: https://github.com/JuliaPlots/RecipesPipeline.jl/pull/95/checks?check_run_id=3426189759#step:8:515.

[t-bltg]

CI is back, re-running checks.

[BioTurboNick]

I believe these errors are only happening because the compensatory change in Plots.jl hasn't happened. The exception is still being thrown from inside Plots.jl, which must be swallowed upstream somewhere, leading to a fallback of some sort.

So, I restored the try/catch block but kept the check for nothing. I added a comment that the block can be removed once Plots.jl makes the compensatory change.

[BeastyBlacksmith]

I always wanted to get rid of that, thanks for digging in!

[BioTurboNick]

Great, thanks all! I'd like to note on the Plots.jl PR what release this will likely be in. Would it likely be v0.3.5, since it isn't breaking? Or are there other breaking changes in the works that would make it v0.4?

[t-bltg]

Would that be ok if we wait for #94 to be finalized / merged before bumping to 0.3.5 ?

Changed my mind about that, let's finalize JuliaPlots/Plots.jl#3765.

[t-bltg]

@BioTurboNick, JuliaPlots/Plots.jl#3765 is emrge: when JuliaRegistries/General#43601 will be merged, you can make a new PR to remove the try/catch part (per #95 (comment)).