[Enhanced Input] The action instance data is incorrectly reset when you call "Add mapping context" on the same mapping multiple times

This is somewhat related to [Link Removed] and was brought up by a [Link Removed]

Setting the CVar input.bRespectIMCPriortyForTriggers to false resolves the issue, but would re-introduce [Link Removed]. 

The additional "Add Mapping Context" call triggers a call to `IEnhancedInputSubsystemInterface::ReorderMappings`, which is where we are retaining the trigger state if needed, which is correct. The issue appears because the ActionInstanceData for that data is cleared though, so on the next tick it will re-evaluate the trigger state and fire the "triggered" event again. 

[Image Removed]

[Image Removed]

1. Create an Input Mapping Context, "IMC_Test"
2. Create an Input Action "IA_Test"
3. Bind "IA_Test" to "M" in the IMC_Test, and add a "pressed" trigger to it. 
4. Add the mapping context to a player controller on begin play
5. Add a BP event listener for the "IA_Test".
   1. Connect a "Flip-Flop" node to add/remove the IMC_Test respectively.
   2. Add Print statements after each, 
6. PIE
7. Press M, and hold it down
8. You will see "add" and "remove" printed. 

Expected: You only see "Add" printed