Added the XIntegrationControl ( re #2134 ) and removed a few compiler warnings in MantidPlot ( re #2818 ).

Refs #2818. Add a few warning options on gcc as there are some that show up on Windows that are useful, in particular conversions that lead to loss of data. 

Declare attribute:

      const NeutronAtom& neutron;

but construct it using from temporary local variable:

const NeutronAtom getNeutronNoExceptions(const uint16_t z, const uint16_t a)
{
  try {
    return getNeutronAtom(z, a);
  } catch (std::runtime_error & ) {
    return NeutronAtom(z, a,
		       NAN, NAN, NAN, NAN,
		       NAN, NAN, NAN, NAN); // set to junk value
  }
}

Atom::Atom(const string& symbol, const uint16_t z, const uint16_t a,
           const double abundance, const double mass, const double density) :
           symbol(symbol), z_number(z), a_number(a), abundance(abundance),
           mass(mass), mass_density(density),
	   number_density(density * N_A / mass),
	   neutron(getNeutronNoExceptions(z, a))
{
}

Changed local attribute 'neutron' to be NeutronAtom instead of NeutronAtom&.

re #2818

Refs #2818. First round of commits for warning clearance on MSVC. There was a macro clash with muParser so I added a wrapper and replaced all includes with this one.

Automatically Rebin the SANS script function TransWorkspace() workspace and commit code that's stopping me from updating re #2794

Refs #2340. Minor fix for python framework if numpy isn't available (in debug on windows at the moment).

re #2843 Should fix failing test buid on windows. Cannot have a using declaration to a nested type with VS compiler.

Add the ability to specify a different cal file for the detector info algorithm for direct inelastic  convert to energy. Re #2837

Fixes #2349. Add a macro that can be used to stop compiler warnings for unused function arguments. Saves every person from doing (void)x //Silence warning. The macro name, UNUSED_ARG, should make it obvious what is happening. Kernel instances have been updated to use it.

Fixes #2839. Make the error message from the Load algorithm much clearer if it can't find an appropriate loader.

Refs #2838: Peak object is pretty complete; some of the conversions to DSpacing and such should be checked to be sure. Tests are in place.