Implement vm.record cheat code

When testing an ercx test with the kevm foundry integration, the proof ended because it reached an unimplemented cheat code vm.record. The documentation for record can be found here.

As a suggestion, you can try this implementation:

1. extending the configuration in foundry.md to add these cells:

        <recordAccess>
          <recordActive> false </recordActive>
           <records>
             <record multiplicity="*" type="Map">
               <recordKey> 0 </recordKey>
               <reads> .List </reads>
               <writes> .List </writes>
             </record>
           </records>
        </recordAccess>

2. add a #call_foundry rule that intercepts the call of the cheat code

   rule [foundry.call.record]:
        <k> #call_foundry SELECTOR _ => #enableRecord ... </k>
     requires SELECTOR ==Int selector ( "record()" )

Here, #enableRecord will start recording the storage operations.

    rule <k> #enableRecord => . ... </k>
         <recordAccess>
            <recordActive> _ => true </recordActive>
            ...
         </recordAccess>

3. overwrite rules for SLOAD and SSTORE to write in the records cells while recordActive is true.

   rule [foundry.recordSLOAD]:
        <k> SLOAD INDEX => #recordRead INDEX ~> #pauseRecord ~> SLOAD INDEX ~> #resumeRecord ... </k>
        <recordAccess> <recordActive> true </recordActive> ... </recordAccess>
     [priority(39)]

   rule [foundry.recordSSTORE]:
        <k> SSTORE INDEX NEW => #recordRead INDEX ~> #recordWrite INDEX ~> #pauseRecord ~> SSTORE INDEX NEW ~> #resumeRecord ... </k>
        <recordAccess> <recordActive> true </recordActive> ... </recordAccess>
     [priority(39)]

Here, #pauseRecord and #unpauseRecord could be used to allow the original SLOAD/SSTORE rules from evm.md to be executed by disabling/reenabling the recording.

    rule <k> #resumeRecord => . ... </k>
         <recordAccess> <recordActive> _ => true </recordActive> ... </recordAccess>

    rule <k> #pauseRecord => . ... </k>
         <recordAccess> <recordActive> _ => false </recordActive> ... </recordAccess>

And #recordRead and #recordWrite can record the operations.

    rule <k> #recordRead V => . </k>
         <id> KEY </id>
         <recordAccess>
           <record>
             <recordKey> KEY </recordKey>
             <reads> R => R ListItem(#buf(32, V)) </reads>
             ...
           </record>
           ...
         </recordAccess>

    rule <k> #recordRead V => . </k>
         <id> KEY </id>
         <recordAccess>
            <records>
               ( .Bag
               => <record>
                   <recordKey> KEY </recordKey>
                   <reads> ListItem(#buf(32, V)) </reads>
                   ...
                 </record>
               )
            </records>
            ...
         </recordAccess> [owise]

    rule <k> #recordWrite V => . </k>
         <id> KEY </id>
         <recordAccess>
           <record>
             <recordKey> KEY </recordKey>
             <writes> R => R ListItem(#buf(32, V)) </writes>
             ...
           </record>
           ...
         </recordAccess>

This cheat code goes hand in hand with vm.accesses, so probably this needs to be implemented as well.

Start with foundry.md and the existing EVM rules in evm.md, then compare the linked Foundry documentation for vm.record and vm.accesses. Implement the record state, cheat-code interception, and SLOAD/SSTORE recording described in the issue; done means the ercx Foundry integration no longer stops at vm.record and the related access data is captured.