connecting an ADC MCP3208 via SPI to RPi is not a problem. Connecting this to scratch depends on the capabilities of a client software. In my school workshop, I use my own implementation for this named 'scratchClient', written in python. I use MCP3202 there. See my website for details http://heppg.de/ikg/wordpress/?p=134
I guess the problem is timing. The force is applied during quite short time, and you need to measure values at double the rate minimum. Assume the 'impact time' is 100 ms, you need to measure 20 times a second or faster. This is beyond the capabilities of scratch.
What you can do is to place the polling task to the scratchClient software, which performs peak detection and then sends a 'hit'-broadcast event to scratch. My scratchClient implementation has a 'adapter' concept which easily allows this.
Another possibility would be to have the target fixed to a pendulum, which the ball starts to move. Place a potentiometer on the pendulum axis. The movement of the pendulum is quite slow, with a board of 70cm length I got 1.4 sec pendulum period time. This is well in the range of scratch response time. And use a heavy, sand filled, soft ball which gives good impulse transfer.
Or use a forked light barrier.
The next problem is to provide a suitable brake or attenuation, otherwise you need to wait for minutes till the next event is possible. But this is mechanics and not scratch then.
Another possibility is to use an electronic analog comparator and a set/reset flipflop. A gpio input and output needed, op amp and cd4013. The pulse peak from FSR sets the FF, when scratch has read this it resets the FF.