cannot create multiple non-revertible update requests to the same public storage slot in a single TX #4736

Historically, if a user wrote to the same public storage slot multiple times in the same transaction, writes would be squashed such that only the most recent write would persist. This was accomplished by:

• the simulator maintaining a map of accumulated storage update requests (see ContractStorageActionsCollector)
• the outputs from the public kernel circuits being modified in TS to squash writes (see removeRedundantPublicDataWrites)

Note, these writes ought to be squashed by the public kernel itself. The reason this is possible without the squashing via the public kernel is that the base rollup circuit trusts that the writes it is provided are valid

However, with the introduction of non-revertible side effects, there are a few pieces of work needed to be able to write to the same public storage slot multiple times in a non-revertible manner.

1. Expand removeRedundantPublicDataWrites to be squash writes from the non-revertible set
2. When recombining revertible and non-revertible writes, also squash redundant writes