Pcbnew: fix issue when creating new tracks: sometimes a ratsnest remains active although a track exists.

  * changed all the RotatePoint() functions in trigo.{h,cpp} to take a double as the
    angle, which is still in tenths of degrees for now.
  * EDA_TEXT::m_Orient is now double, still in tenths of degrees (for a short while),
    but the double means parsing files and saving them use a different format string.
++pcbnew
  * DRAWSEGMENT::GetStart() and GetEnd() do not operate for S_ARC like they used to.
    They are now simply accessors for m_Start and m_End.  Use DRAWSEGMENT::GetArcStart()
    and GetArcEnd() and GetCenter() for arcs.  specctra_export.cpp was the only
    source file dependent on the old behavior.
  * DIMENSION::m_Text is now contained, not dynamically allocated.
  * more kicad_plugin work.

* Enable replace toolbar button and menu entry.
* Improve find replace logging granularity.
* Fix find and replace dialog control hiding and disabling.
* Minor improvements to the SCH_FIND_REPLACE_DATA object.
* Move find collector list iterator into the collector object.
* Add visibility override flag to EDA_ITEM to temporarily show items that
  are hidden during find and replace.

  * added BOARD::GetPad(int), made BOARD::m_NetInfo private, and contained
    within the BOARD, not dynamically allocated and owned via pointer.
  * added BOARD::GetPadCount()
  * changed NETINFO_LIST::GetCount() to GetNetCount()
  * added BOARD::GetNetCount()
  * more kicad_plugin work.

Pcbnew: fix bad initialization of default track width and vias sizes when starting pcbnew with no board loaded.

 * Fix compil issue about automatically created pcb_plot_params.h :
     pcb_plot_params.* moved from pcbnew to common because item_io.cpp (using this file) is compiled in common.

In common.cpp GetTimeStamp is renamed GetNewTimeStamp (a better name).
Pcbnew: prepare work to calculate connections between pads that inteserct and therefore can be connected without any track (composite pads).

which should eventually go away in favor of kicad_plugin.cpp.
Add CMake options USE_NEW_PCBNEW_LOAD and USE_NEW_PCBNEW_SAVE, which should not be
used yet, and both default to OFF accordingly.

and TEXTE_MODULE size in kicad_plugin.
Link in io_mgr and kicad_plugin but do not call them yet

  * Removed Pcb_Frame argument from BOARD() constructor, since it precludes
    having a BOARD being edited by more than one editor, it was a bad design.
    And this meant removing m_PcbFrame from BOARD.
  * removed BOARD::SetWindowFrame(), since BOARD::m_PcbFrame gone.
  * Removed the global BOARD_DESIGN_SETTINGS which was in class_board.cpp
  * Added BOARD_DESIGN_SETTINGS to the BOARD class, a full instance
  * A couple of dialogs now only change BOARD_DESIGN_SETTINGS when OK is pressed,
    such as dialog_mask_clearance, dialog_drc, etc.
  * Removed common/pcbcommon.cpp's int g_CurrentVersionPCB = 1 and replaced it
    with build_version.h's #define BOARD_FILE_VERSION, although there may be a
    better place for this constant.
  * Made the public functions in PARAM_CFG_ARRAY be type const.
    void SaveParam(..) const and void ReadParam(..) const
  * PARAM_CFG_BASE now has virtual destructor since we have various ways of
    destroying the derived classes and boost::ptr_vector must be told about this.
  * Pass const PARAM_CFG_ARRAY& instead of PARAM_CFG_ARRAY so that we can use
    an automatic PARAM_CFG_ARRAY which is on the stack.
  * PCB_EDIT_FRAME::GetProjectFileParameters() may no longer cache the array,
    since it has to access the current BOARD and the BOARD can change.
    Remember BOARD_DESIGN_SETTINGS are now in the BOARD.
  * Made the m_BoundingBox member private, this was a brutally hard task,
    and indicative of the lack of commitment to accessors and object oriented
    design on the part of KiCad developers.  We must do better.
    Added BOARD::GetBoundingBox, SetBoundingBox(), ComputeBoundingBox().
  * Added PCB_BASE_FRAME::GetBoardBoundingBox() which calls BOARD::ComputeBoundingBox()

  * Removed Pcb_Frame argument from BOARD() constructor, since it precludes
    having a BOARD being edited by more than one editor, it was a bad design.
    And this meant removing m_PcbFrame from BOARD.
  * removed BOARD::SetWindowFrame(), and BOARD::m_PcbFrame
  * Removed the global BOARD_DESIGN_SETTINGS which was in class_board.cpp
  * added BOARD_DESIGN_SETTINGS to the BOARD class, a full instance
  * a couple dialogs now only change BOARD_DESIGN_SETTINGS when OK is pressed,
    such as dialog_mask_clearance, dialog_drc, etc.
  * Removed common/pcbcommon.cpp's int g_CurrentVersionPCB = 1 and replaced it
    with build_version.h's #define BOARD_FILE_VERSION, although there may be a
    better place for this constant.
  * Made the public functions in PARAM_CFG_ARRAY be type const.
    void SaveParam(..) const and void ReadParam(..) const
  * PARAM_CFG_BASE now has virtual destructor since we have various way of
    destroying the derived class and boost::ptr_vector must be told about this.
  * Pass const PARAM_CFG_ARRAY& instead of PARAM_CFG_ARRAY so that we can use
    an automatic PARAM_CFG_ARRAY which is on the stack.\
  * PCB_EDIT_FRAME::GetProjectFileParameters() may no longer cache the array,
    since it has to access the current BOARD and the BOARD can change.
    Remember BOARD_DESIGN_SETTINGS are now in the BOARD.
  * Made the m_BoundingBox member private, this was a brutally hard task,
    and indicative of the lack of commitment to accessors and object oriented
    design on the part of KiCad developers.  We must do better.
    Added BOARD::GetBoundingBox, SetBoundingBox(), ComputeBoundingBox().
  * Added PCB_BASE_FRAME::GetBoardBoundingBox() which calls BOARD::ComputeBoundingBox()

    Enhanced algorithms to calculate board connections:
    Previously, a track was seen connected to a pad only if the track end is
    exactly on the pad position.
    Now a track is seen connected to a pad if the track end is inside the pad shape.
    Algorithm to calculate pads connections to track is still very fast.
    However some other functions (drag pads, track len calculation ...)
    still need the track end exactly on the pad position.
Dead code removal.