ALFA_SensitiveDetector can make hits with plate==100 for GVS volumes.
Usually 0<plate<10. The hit plotting code in ALFA_SimHitsTestTool assumes
a maximum value of 10 when it uses plate as the index to arrays, so when a
hit with plate==100 is processed this ends up well outside the array and
corrupts the stack. The fix is to put the GVS hits into a separate histogram,
skipping the stage where they are used as the index to an array.
Thanks to @ssnyder for finding the issue.