docs: remove legacy l2-on-l2 demo

docs/l2-on-l2.md

-# Running Cannon on L2
-
-The original Cannon prototype allows challenging L1 blocks on L1. Normally, the
-challenger should always fail, as L1 blocks are valid by virtue of being
-included on-chain.
-
-The next milestone is to allow challenging L2 blocks on L2. What this proves is
-that the extra logic we added in
-[l2geth](https://github.com/ethereum-optimism/reference-optimistic-geth) (aka
-reference-optimism-geth) does not break anything. It's also a good way to
-exercise our devnet infrastructure.
-
-Running the Cannon demo on a mainnet (L1) fork is a simple as doing:
-
-```bash
-# from repo root
-demo/challenge_simple.sh
-# or
-demo/challenge_fault.sh
-```
-
-For L2, you'll need first to run the devnet locally. For this, clone the
-`develop` branch of the [optimism
-monorepo](https://github.com/ethereum-optimism/optimism), then run:
-
-```bash
-yarn && make build && make devnet-clean && make devnet-up
-```
-
-If you're having trouble building, here's a [full
-transcript](https://github.com/ethereum-optimism/cannon/wiki/Bedrock-Full-Devnet-Setup)
-of all the commands required to run on a fresh cloud linux machine.
-
-Note it's important to run `make devnet-clean` before each invocation of `make
-devnet-up` to work around some issues at the time of writing.
-
-Then you can run the L2 demos:
-
-```bash
-# from repo root
-demo/l2_challenge_simple.sh
-# or
-demo/l2_challenge_fault.sh
-```