eth_call fails when batch has more than 1 tx for the same acc

[tclemos]

The way the node process unsigned transactions is by getting the l2 block, then identifying the batch related to this block in order to set the state root we must use when processing the transaction. The code always uses the state root of the latest l2 block related to the batch, so depending on the order of the transactions we might have an false ProviderError: nonce intrinsic error

Here is how it works:

Imagine a scenario like this:

batch 1
block 1 acc 1 tx 1 nonce 1
block 2 acc 2 tx 1 nonce 1
block 3 acc 1 tx 2 nonce 2

if I do a call for block 2, the nonce of the acc 1 should be 1... but since we have the block 3 attached to batch 1, we use the state root of block 3 instead of the root of block 2... and this causes the problem