Wedge in some more SLEEF functions - this is most inelegantly done

I am doubtful about the SLEEF allocators

Tidy, and remove the very approximate atan2f

Apply actual SLEEF SIMD code for -to-polar

Increase alignment in allocate

More meaningful timing code

Some tidying, + make simple dispatch functions inline

Add some basic SLEEF support - I don't think any of this is useful as it stands though as it takes no advantage of the actual vector functions yet

More meaningful timing code

Some tidying, + make simple dispatch functions inline

Explicit init

Fix incorrect precision in checks for some builds

Fix ipp divide_to

Tests and vDSP implementation for complex divide/dot

Add complex vector divide and dot

Allow zero-initialisation for things for which that doesn't necessarily mean zero

Add min, max, rms

Merge

Do the proper thing with Clang/VC

No longer using Travis