TSSIM/JESIM v46 - Mercator scale / km-per-cell calibration fix

Problem:
  v45 used g_map_top_lat_deg=42.25 to derive map scale. That produced:
    projected km/cell ~= 3.39
    ground km/cell at map center ~= 2.75

  Fitting the uploaded station latitude/longitude table to the station grid
  coordinates in tsunami_data.bin gives:
    projected km/cell ~= 5.72096
    ground km/cell at map center ~= 4.63

  The old scale was about 0.59x of the calibrated scale, so fault stamp
  down-dip depth increments were too shallow and source-node fault area was
  too small. The area effect is about (4.63/2.75)^2 ~= 2.85, which is roughly
  Mw +0.30, matching the observed M7.5-class low bias.

Changes:
  - Calibrated Mercator constants:
      center lat = 35.91325
      center lon = 135.79685
      projected km/cell = 5.72096
      top-lat display reference ~= 46.29N
  - map_projected_km_per_cell_from_top_lat() now returns the calibrated value.
  - v45 stamp plane-depth behavior is kept, but now uses the corrected km/cell.
  - Existing v45 features are preserved.

Build:
  build_win_mingw.bat
  python jesim_ws_relay.py
