feat(client): Channel Timeout Tests

[refcell]

Description

#427 implements an optimization for the client program to avoid needing to walk back the full chainnel_timeout window of L1 blocks, but this codepath isn't tested.

This issue is to write a test suite that incorporates the client's driver logic that walks back the channel_timeout window.

The first test case to cover is that a partial channel causes the client to fail. This currently doesn't work since when witness data is missing, the client is stuck in an infinite loop waiting for the oracle to retrieve the data.

A comprehensive test case to cover the happy path should do the following.

• Start derivation at the L1 origin of the agreed upon pre-state optimistically.
• The first channel that is closed is considered partially read, causing decoding to fail.
• The pipeline should bubble up a batch decoding failure error.
• Then the client's driver should reset the pipeline and re-start channel_timeout L1 blocks prior to the L1 origin.

[refcell]

We should be able to use derivation test fixtures from https://github.com/ethereum-optimism/tests