Alexey Stukalov

I've just tried and with the current `plotly` version the subplot margins issue is still there (margins too wide, subplots have uneven sizes), so I've rebased the PR.

For my plot I needed both the interactive web version and the PDF one (so I have to use ggplot2/ggplotly, since AFAIK the free plotly.js version does not support exporting...

Note that there is no regression in coverage, the PR is just 2 lines of code less.

@pkofod gentle ping. :) could you please review this PR? It should provide a nice performance improvement for Particle Swarm.

Unfortunately, the original problem would be a bit cumbersome to share. But I think it should be easily reproducible with any problem that have >100 parameters. In my case the...

@pkofod Yes, preallocation should help for bigger problems. For my PR I wanted just the minimal changes that speedup linear algebra. I can try preallocation too, if *Optim.jl* has preallocation-enabled...

Now these functions should be fixed and properly tested. The latter would require merging JuliaDiff/DiffTests.jl#11 first and updating the deps here.

@mcabbott @fredrikekre @devmotion I've refactored the code so that in-place ldiv/mul are also supported. These changes should now be covered by tests with the updated *DiffTests* package. There are some...

> @alyst, this PR is out of date. `LambertW` has changed in the meantime. I (or we) need to merge the changes into this PR branch. Let me know how...

> I'm not completely against moving LambertW.jl, but I'd like to see arguments for why. * it's more likely that packages like *SpecialFunctions.jl* will have (co-)maintainers in the long run...