basic implementation of NIntegrate

[mmatera]

Here it is a basic implementation of NIntegrate, working both on intervals and d-dimensional regions.
If scipy.integrate is available, uses quad as a default integrator, and allows to call other routines (romberg, and nquad, but others can be included later). Otherwise, or if these routines fails, a native Simpson's rule adaptative routine is included, and set as default if scipy is not available.
With #1161, I expect that this also works for complex valued functions.

[rocky]

Overall I like this! Thanks!

On additional suggestion is to change CompiledCode#str to report which variation of compilation was used now that several methods are tried.

[rocky]

When I assign this to a variable and run that on anythng anything other than 0 as the b argument this hangs. I am guessing this definition isn't quite right?

[rocky]

One other thing I note is that when I look compilation.py pyflakes is reporting a number of small lint notes involving unused variables and imports.

[mmatera]

Yes to both things: we need at some point to review the implementation of Compile in a closer way. However, this is a task for a couple of weekends. What motivated me to dig into this right now is that I needed this work for the NIntegrate implementation.

[rocky]

I like this and it is much needed. Thanks!

As suggestion for improvement if you are up for it is to add in CompiledCode#__str__ an indication of what kind of compiled code we have now that there are several methods.

[mmatera]

I like this and it is much needed. Thanks!

As suggestion for improvement if you are up for it is to add in CompiledCode#__str__ an indication of what kind of compiled code we have now that there are several methods.

Thanks! This last commit should attend to your suggestion.