Enable option /fp:precise explicitly in Visual Studio projects

Unlike VC++, Intel compiler by default performs non-safe optimization of floating-point arithmetic (/fp:fast). This leads to multiple regressions in OCCT tests (built in Release mode with optimizations enabled).

The standard procedure to use Intel compiler (to be employed by OCCT) is to switch existing VS project to Intel compiler, with option integrated to VS.

In order to ensure that OCCT projects converted in this way work as expected we need to enable option /fp:precise explicitly in Visual Studio projects generated by WOK.

Note that for 32-bit Intel compiler use of /fp:precise leads to poor performance (about 20% slow down for OCCT); this can be repaired by setting FP model option to /fp:source (in VS project Properties it is on separate tab "Code Generation [Intel C++]"). This is to be done manually.