Class GuiSizesHelper has tons of bad code

[johsin18]

Class org.openstreetmap.josm.tools.GuiSizesHelper IMHO has tons of bad code

• literal "96f" is repeated many times, sometimes with "f" suffix (float), sometimes without (int)
  • Mitigation: at least make it a named constant
  • better Mitigation: keep the pixel density as a member, remove 96f everywhere (also see below)
• Toolkit.getDefaultToolkit().getScreenResolution() never returns anything else than 96, at least on Windows, even when scaling!=100% is set for screens
  • in case it returns 72 for MacOS, we are in even more trouble
  • Mitigation: getScreenResolution should never be called, just read the "gui.scale" preference
  • getScreenResolution() tries to achieve something similar to HiDPIsupport.getHiDPIScale(), but in a different, possibly conflicting way
• GuiSizesHelper is not a very descriptive name
  • Mitigation: rename it to UserPreferredUiScaling or something alike, as the actual screen resolution does not play a role
• method isHiDPI() is not used at all
  • Mitigation: remove method, together with getPixelDensity(), which is only used by isHiDPI()
• methods check everywhere that passed dimensions are >0. However, for values 0, the computations should result in exactly the same, so the check is not needed. If negative values are passed, this is proabaly a sign for problems in the calling code.
  • Mitigation: remove the checks
• Most GUI elements are scaled, but not the mouse cursors.
  • Mitigation: rather rely on Toolkit.getBestCursorSize()
• There is no test coverage for its functionality at all
  • Mitigation: add tests
• Adds an indirect dependency to Config for all using classes/tests

Bottom line: Before fixing this very broken class, we should probably remove it together with its functionality, as already proposed in https://josm.openstreetmap.de/wiki/Help/HiDPISupport (at the very bottom), and rather fix remaning bugs in the official HiDPI support.

SCNR I volunteer to get this done.