Algorithm is still the same... now need to know what it really meant to do

*)  Implement a framework for "Data Load On Demand".

*)  Implement FP_LIB_TABLE* PROJECT::PcbFootprintLibs(), which is the first
    prototype.

This allows the project specific footprint tables to be part of the Module Editor
when invoked from Eeschema.

Fix an annoying behavior of new quasi modal frames  on Windows (when closing them, the caller frame was no more on top )

fix also a bug in python script round_value_robin.py.

*) Eeschema can now show the footprint editor.

*) Eeschema can now invoke the footprint picker from the library part field editor.

*) KIWAY_PLAYER::ShowModal() takes aResultantFocusWindow that tells what window
   to give the focus to.  Required since frames are often near the top of the
   hierarchy and they are invoked by a peer, not a parent.

Kicad: fix crash on exit, when Libedit was called during a session. It fixes also a crash if Libedit was called, closed and reopened.
Happens only if eeschema was called from the Kicad manager.

footprint editor: fix crashes. In this fix, I removed the assumption the parent frame is the board editor.
However, this assumption is still present here and there  in the moduleframe code.

wx 2.8 now builds, but the quasi-modal support seems broken on 2.8.  Wanted others to be able to build at least.

*) KIWAY_PLAYER::IsModal() is now a retained state, controlled by SetModal()

*) Fully re-work the KIWAY_PLAYER::ShowModal() to use a nested event loop.

*) Add support to DIALOG_SHIM for a "quasi-modal" dialog presentation and mode.
   See top of dialog_shim.cpp about that for benefits and need.

*) You can now pick footprint from the schematic component field dialog, although
   if you do this before you open the BOARD, you will only get the global footprint
   libraries, not also the project specific ones.  Opening the BOARD first avoids this
   problem.

This is the first example of cross KIFACE invocation, it is also the first
instance of using a TOP_FRAME other than FRAME_PCB as the first thing. It works,
but it's missing support for opening the project specific table because
historically the FRAME_PCB did that. This is now starting to expose all the near
term needs for KIWAY_PLAYER <-> PROJECT interaction, independence and out of
sequence usage.

A fix for this will be coming in a few days.

However it mostly starts to show why the KIWAY is terribly useful and important.

In particular HitTest for zones *do not* select the nearest vertex/edge as a side effect

*) Rework the set language support, simplify it by using KIWAY.  Now any major
   frame with a "change language" menu can change the language for all KIWAY_PLAYERs
   in the whole KIWAY.  Multiple KIWAYs are not supported yet.

*) Simplify "modal wxFrame" support, and add that support exclusively to
   KIWAY_PLAYER where it is inherited by all derivatives.  The function
   KIWAY_PLAYER::ShowModal() is in the vtable and so is cross module capable.

*) Remove the requirements and assumptions that the wxFrame hierarchy always
   had PCB_EDIT_FRAME and SCH_EDIT_FRAME as immediate parents of their viewers
   and editors. This is no longer the case, nor required.

*) Use KIWAY::Player() everywhere to make KIWAY_PLAYERs, this registers the
   KIWAY_PLAYER within the KIWAY and makes it very easy to find an open frame
   quickly.  It also gives control to the KIWAY as to frame hierarchical
   relationships.

*) Change single_top to use the KIWAY for loading a KIFACE and instantiating
   the single KIWAY_PLAYER, see bullet immediately above.

*) Add KIWAY::OnKiwayEnd() and call it from PGM_BASE at program termination, this
   gives the KIFACEs a chance to save their final configuration dope to disk.

*) Add dedicated FRAME_T's for the modal frames, so m_Ident can be tested and
   these modal frames are distinctly different than their non-modal equivalents.
   KIWAY_PLAYER::IsModal() is !not! a valid test during the wxFrame's constructor,
   so this is another important reason for having a dedicated FRAME_T for each
   modal wxFrame.

On balance, more lines were deleted than were added to achieve all this.

Eeschema: fix unwanted creation of noname.pro (or other useless config file), when opening eeschema. Fix also an issue when saving the config (from save config menu) which saved the config file in an unknow folder.

- Converted the Next/Prev C casts to static casts and removed the type
  unsafe ones
- Splitted as virtual the VIA::Flip member instead of using RTTI
- Heavily refactored the 'unconnected track' cleanup routine
- Misc constification

Plots function: fix a bug about virtual PLOTTER::Text, which was not virtual for derived classes due to a missing parameter in ::Text in these classes. Noticeable only in SVG plot.
SVG plot, fix a missing reinitialization in plot lines, which could define a filled polyline, instead of a simple polyline
(these fixes solve Bug #1313084 )

Reverse Marco's patch, pending a proper review of my last patch from an English speaking Mac developer.

[MacOSX] Fix for name collision with OSX headers, Maciej, Tomasz or others feel free to rework if you want/like

3D viewer: Commit patch to enable the rendering of wrl files generated with Meshlab. See https://bugs.launchpad.net/kicad/+bug/804778

ENDPOINT_T enum type

eeschema: fix the very minor bug  #1298094. But it was due to a bad bounding box calculation, which is now fixed.

(just replaced it with a member)