Allow curve_fit() to take a data argument to be passed to model function

[olehmer]

When using the curve fit function I found the need to pass constant arguments to the model function. Using global variables was doable but unfortunate. The margs=[] argument will work with all legacy code using the curve_fit() function but allow future calls to handle a wider range of use cases.

[mlubin]

Thanks for taking the time to put together this contribution, but this issue has been discussed at length:
JuliaNLSolvers/Optim.jl#102
JuliaNLSolvers/Optim.jl#53
JuliaNLSolvers/Optim.jl#94
JuliaNLSolvers/Optim.jl#85
and the conclusion is to use Julia's closure functionality, as described at:
https://github.com/JuliaOpt/Optim.jl#optimizing-functions-that-depend-on-constant-parameters
Closures have no performance penalty in Julia 0.5 and later.

[ChrisRackauckas]

You could also use call overloading and have model be a type with a field for the data. That would allow you have the data "explicit" and accessible (unlike a closure), but also requires no API change. It's how parameters for differential equations are handled in JuliaDiffEq (ParameterizedFunctions.jl)

[blakejohnson]

I agree with @mlubin, I think closures are a much better way to handle this scenario, especially now that closures have no overhead in julia 0.5.