Benchmark fraud proof validation in pre_dispatch

[NingLin-P]

Currently, the weight of the submit_fraud_proof extrinsic only covers the handling part of the fraud proof (i.e. prune bad ER, slash submitter and revert head receipt number, etc) while the fraud proof verification is also expensive compared to other normal extrinsic, thus we should benchmark the fraud proof verification in pre_dispatch and add the resulting weight to the submit_fraud_proof extrinsic.

However, there is an obstacle because the fraud proof verification involves the host function for querying block history and old state, etc, from the client side, which is hard to simulate in the benchmark, and also the current benchmark infra is more about benchmarking the runtime.